DO ENVIRONMENTAL FACTORS ALTER USER’S BEHAVIOR: EVIDENCED IN MOVEABLE FURNITURE ON A UNIVERSITY CAMPUS? By Rachel Rose Wilke A THESIS Michigan State University in partial fulfillment of the requirements Submitted to for the degree of Environmental Design─Master of Arts 2019 ABSTRACT DO ENVIRONMENTAL FACTORS ALTER USER’S BEHAVIOR: EVIDENCED IN MOVEABLE FURNITURE ON A UNIVERSITY CAMPUS? By Rachel Rose Wilke Major studies have shown that seating is an important factor in explaining popularity or vacancies of plazas and other forms of public spaces that support public gatherings and social interactions. Although many studies have researched how much seating is needed, few studies to date examine the impact that movable types of seating have on people’s behavior. This research focuses on the response of people’s behavior towards moveable furniture within plazas. Specifically, it explores this understudied area of landscape design through the use of time-lapse photography to capture the movement of nomadic chairs at a central plaza on Michigan State University’s campus. Over 384 recorded images from this project that were coded to document the number of chairs in distinct zones within the plaza. Regression analysis was then used to examine the relationship between the densities of chairs in particular zones in response to: time of day, weather, and temperature. The analysis suggests that moveable furniture is a tool in exhibiting how these abiotic factors mediate user behavior. The findings of this research show the importance of providing nomadic furniture in plazas and the importance of proposing design guidelines for successful seating placement. Copyright by RACHEL ROSE WILKE 2019 This thesis is dedicated to my mom for always being there when I need support and someone to talk to. iv ACKNOWLEDGMENTS Foremost I would like to share my sincere gratitude for my thesis advisor, Dr. Linda Nubani. She welcomed me on as one of her master’s students when I was in search of a new chair. I would not have excelled in my work without her guidance and mentorship. I would also like to thank my committee members guiding me through the thesis process. Statistics models and the programs would still have been a mystery to me without Dr. Noah Durst spending countless meetings dissecting the meanings and algorithms that lead to my conclusions. I also want to thank Dr. Jun-Hyun Kim for all of his insightful comments on my work as a student leader and now as a master’s student. I would also like to acknowledge Dr. Pat Crawford who first introduced me to my thesis topic through her connections with MSU Landscape Services (Infrastructure Planning and Facilities). This brings me to send my sincere thanks to MSU Landscape Services, which my study would not have been possible without their help bringing the chairs in and out every day. They also supplied the time lapse cameras that collected all of my photos I used for my analysis. I thank the offices of the religious Studies in Wells Hall and the second floor office of the International Center for allowing me to place my cameras in their windows and bother them every time I came in to download the pictures and supply new batteries. I also want to send my thanks to CSTAT that helped me with part of my statistical analysis. They spent time with me so I could understand the different ways in which I can test my data. v Last but not least, I would like to thank my family and all my friends that have supported me along this process. vi TABLE OF CONTENTS LIST OF TABLES ………....…………………………………………………………….……. ix LIST OF FIGURES ………………………………………………………………………….... x 1 CHAPTER ONE ……...……………………………………………………………………….. 1 INTRODUCTION ...………………………………………………………………………… 1 Preamble ….…………………………....…………………………………………......….... 1 1.1 Importance of Seating in Public Spaces ……………………………………………….. 4 1.2 Research Outline ..…………………………………………………………………...… 4 1.3 Research Objectives and Hypothesis ...…...…………………...……………………..... 7 CHAPTER TWO ...………………………………………………………………………..…... 7 LITERATURE REVIEW ………………………………………………………………....… 7 Preamble ...………………………………………………………………………………... 7 2.1 Public Plazas ……………….…………………..…………………………............…… 9 2.2 Urban Plaza Design Elements …………….……………………………………............ 10 2.3 Human Behavior ……………….……………………………………….............……... 12 2.4 Mapping ………………………….……………………………………………....……. 13 2.5 Design Response to Climate Instability ...…………………………..……………..…... 14 CHAPTER THREE ...……………………...………………………………………….………. 14 METHODOLOGY …………………………………………………………………...…....... 14 Preamble ...…………………………………………………..…………………………….. 16 3.1 Furniture ..………………………………………………...…………………...…….…. 17 3.2 Site ………………..…………………………………..……….………….…................ 22 3.3 Independent Variables ...……………………………………….……..…….……….… 24 3.4 Data Collection …..…………………………………….…..……………...……….….. 27 3.5 Visual Analysis ……………….……………………..…………………...………….… 29 3.6 Statistical Methods ………………………………………..…………………………... 35 CHAPTER FOUR ...…………………………………..………………………….……………. 35 FINDINGS AND DISCUSSION …………………………………….…………………........ 35 4.1 Zone Summaries ...…….………………………………….…………..……….………. 36 4.2 Initial Trends ………………………..……………………......….…………………...... 39 4.3 Interaction Models ………………………………...…………..……………………..... 54 4.4 Final Model ...…………………………………………………………..……...………. 60 CHAPTER FIVE ...……………………………………………………………………………. 60 CONCLUSION ……………..………………………………………………………...…....... 62 5.1 Design Implications .........………………………………………………….......…..….. 63 5.2 Limitations …………………………………..………………………..……...…….….. 5.3 Future Research ………………….…………...……………………………..………… 65 vii BIBLIOGRAPHY …………………………………………………………….…….…………..66 viii LIST OF TABLES Table 1. Summary of missing data ...….………...……………………………………………… 34 Table 2. Summary of average chair count per zone ……………………………………..……… 36 Table 3. Linear regression for Temperature, zone, and chair count ……………….…………… 41 Table 4. Linear regression for time group, zones, and chair count ………………...…………… 45 Table 5. Linear regression between weather, zones, and chair count ……...…………………… 52 Table 6. Linear regression between temperature, time, weather, zones, and chair count .……… 54 ix LIST OF FIGURES Figure 1. Shows pedagogical exercise in the process of developing proposed hypothesis …… Figure 2. Conceptual Framework for all the Variables in the Study………...……………….... Figure 3. Photo showing Adirondack chairs on site. Source: by Author ……..…..…………... Figure 4. Photo showing Adirondack chairs. Source: by Author……………..……………..… Figure 5. Image showing site of idea chair project. Source: Google Earth ……………..…..… Figure 6. Photo showing Adirondack chairs setup for each morning as a control. Source: Google Earth………………………………………………………………...…………..……... Figure 7. Visualization of the seven zone group locations. Source: by Author……………..…. Figure 8. Distribution of land between the seven zones.............………………….…………… Figure 9. Temperature range through out the study………..…………….……………….…… Figure 10. Locations of the two cameras used to capture images above and below the tree canopy……………………………………………..…………………………………………… Figure 11. (Left) Typical image captured by BRINNO camera. (Right) Image of the specific camera used in study. Source: www.bhphotovideo.com………….........................…………… 5 6 15 15 17 18 20 21 23 26 26 28 Figure 12a. Demonstrates step one in coding chairs. Photo taken from time lapse camera....… 28 Figure 12b. Demonstrates step two in coding chairs. Grid overlaid on picture to set a base for the application of zones .………………………………………………….…………………… Figure 12c. Demonstrates step three in coding chairs. Dots are overlaid on top of chairs. …... Figure 12d. Demonstrates step four in coding chairs. Zones are added. Chairs are tallied for each zone and added to excel sheet..…….….…………………………………………..……… 29 29 Figure 13. Format of excel document before STATA……………………..…….…..………… 32 33 Figure 14. Shows output from STATA ………………………………..…………………….... Figure 15. Scatter plots depicting the relationship between temperature, zones, and chair 37 count……………………………………………………………………………………………. Figure 16. Scatter plot depicting the relationship between time group, zones, and chair count.. x 38 Figure 17. Scatter plot depicting the relationship between weather, zones, and chair count....... 39 Figure 18. Interaction model for Temperature, zones, and chair count………………………….41 Figure 19. Interaction model for time group, zones, and chair count…………………………....45 Figure 20. Interaction model for weather, zones, and chair count ...……………..………….…..52 xi CHAPTER ONE INTRODUCTION Preamble This chapter highlights the importance of seating in public spaces and presents a conceptual framework drawn from literature that demonstrate the variables that impact the use of movable chairs and their location. 1.1 Importance of Seating in Public Spaces Public space is one of the most important types of pedestrian realms; the use of these spaces has always been of interest. Socializing and connecting with your community has occurred in these public spheres and continues to act as a main point of connection (Peters, 2010). Research, starting in the 1980’s with social life projects by renowned researcher William Whyte, found a relationship between the likelihood of social interaction and the design of a space. Design can work to create a popular social space or a picturesque destination that is unused (Montgomery, 1998). The main reason for this difference in the number of users is based on design elements in the site and the configuration of the space. Visitor’s behavior is directly related to the environmental factors they encounter while in a space. Designs therefore have a direct relation to the interpretation, processing, and usability of the space. There are many different components that give a public space a sense of life, the strategic layout and selection of elements can be a determining factor in use. One of the key elements that contribute to either the popularity or vacancy of these social spaces is seating (Whyte, 1980). Users of the space will use benches, chairs, sitting walls, stairs and any linear surface that fits their needs. Seating is a piece of furniture that takes on many different forms and fits many different needs. 1 The way these needs are met, as well as function with in plaza spaces has been researched through behavioral analysis since 1971 when William Whyte, (1980) started his Street Life Project to study these social realms. In his study he examined a small plaza in the heart of the business district in New York. He emphasized that the importance of these public spaces is the ability for the user to see and be seen. The visitors of these plazas populated the space on their lunch hours to relax outside and to socialize. More recent studies continue to examine public spaces with modern mapping techniques such as GIS and GPS software. (Ghavampour, 2016; Ye, 2013). In Ghavampour’s (2016) study, the researchers looked at multiple parks and how the public uses the space with the use of GIS. This modern technology provided the ability to map these elements while grouping them into specific categories, allowing for real time comparisons. Actively trying to improve the designs of plazas has been and will continue to be a topic that is of interest as it directly effects the public. Previous research has examined the different elements that create a space and how these elements are spatially organized. The majority of these studies compare the arrangement of design elements to human behavior. However, it is the translation of these findings into current design guidelines that is lacking within the outcomes of previous research (Adbulkarim & Nasar, 2014). Knowing that this is an aspect of research that is typically not acknowledged, it is important for this proposed research to make a point of translating the findings into real world guidelines. A type of limitation that has shown up in previous research without the use of modern technologies, was the chance of bias influencing data collection from the researcher. Previous studies such as William Whyte’s used methods that relied on observation and inference that could have been influenced by the impressions that they held before conducting the experiment. In his experiment, time lapse photography was implemented for data collection 2 while the analysis was based upon ethnographic descriptions of interactions between users and design elements. This type of analysis can be effected by presumptions held by the researcher; such as subconsciously looking for activity on the edges if they previously believed that more people will congregate in these locations. In this proposed research data was analyzed in a manner that is free from opinion, allowing for unbiased results through a quantitative method based on visual coding. The last major limitation viewed in previous studies was in the way experiments were conducted, which took place in unnatural environments (Ye, 2013). Experiments used to be performed in laboratory like settings that yielded results that may have been affected due to the sterile conditions. However, in current studies like this proposed research, collecting data can take place in existing public spaces, where the participants may be unaware they are being researched which yields result that are true to real life conditions. Due to the diverse roles design plays in our society, moveable furniture is a tool that can work with, instead of against this diversity. Planners and designers are looking to improve the quality of use in public spaces through new information and design application. Psychologists and sociologists look at design and the use of these spaces from a behavioral perspective. Whether you are on the side of design or behavior they all interrelate; they both feed into and off of each other to create a successful space; if one aspect is missing the other falls apart. Nomadic furniture and the study of its relationship to public behavior ties both together. Abdulkarim and Nasar (2014) stated that seats can be a defining factor in the visitability of a space, which is important knowledge in the planning process and application of design elements. Designers will be interested in this information as a design aspect while sociologists will look at this element as a social stimulator. This study is relevant in the changing urban scene and the people that make up these public spaces (Siu, 2015). 3 1.2 Research Outline The central research question of this thesis is: do abiotic factors influence user behavior? The aim of this study was to identify the patterns exhibited in the use of design elements in public spaces. In this study, associations among different variables were identified and used to help address this main research question. The results for this study can be used in similar sites to asses when and where nomadic furniture will work best in a plaza space. The patterns can be summarized into a set of guidelines that will outline seating preference in urban plazas. With guidelines for seating, more informed decisions can be made for their use and placement in current and future works. Figure 9 shows a framework that illustrates these relationships. Chapter two reviews the previous literature on the subject. Chapter three explains the methodology. Chapter four explores the findings and leads through a discussion. Chapter 5 concludes the research and explores the applications. Based upon the research question, three hypothesis were formed in this research. 1.3 Research Objectives and Hypothesis The purpose of this study is to examine whether nomadic furniture plays a role in the behavior of people in urban public spaces. When the public uses a space, few behavioral patterns begin to emerge as they interact with the moveable furniture. Questions to be addressed in this research are: 1)What will be the common behavioral pattern exhibited in the chairs’ movement? 2) How will variables such as climate contribute to the behavioral patterns expressed in the placement of chairs? Specifically, the following three hypothesis were proposed for this study through performing a pedagogical exercise before starting my experiment: 4 Hypothesis 1: Increases in temperature will lead to increases in the density of chairs in shaded zones and decreases in the density of chairs in unshaded zones Hypothesis 2: The morning, afternoon, and evening time groups will show distinct increases in the chair density of certain zones over others. Hypothesis 3: A decrease in the amount of cloud coverage will lead to an increase in the density of chairs in shaded zones and a decrease in density of chairs in more exposed zones. The following graphs document part of the process used in the study to come up with my hypothesis that support my research. Figure 1. Shows pedagogical exercise in the process of developing proposed hypothesis. 5 Figure 2. Conceptual Framework for all the Variables in the Study 6 CHAPTER TWO LITERATURE REVIEW Preamble It may seem expected that people use public spaces, especially in urban areas when private space is limited. However without the use of proper planning and design, these spaces can be rendered vacant. A space that is abundant almost anywhere you go is the urban plaza, a place that can serve many different roles in a community. Every urban plaza is unique and has its own identity, with successful plazas molding themselves around their surroundings (Wang, 2001). Many people have looked into and studied human behavior in response to the elements that make up an urban space to better understand how they function and how to better design them (Whyte, 1980; Adbulkarim & Nasar, 2014). However, there is limited research examining the arrangement of these elements and their resulting impact on human behavior. 2.1 Public Plazas William Whyte, a renowned American urbanist, pioneered the examination of human behavior within public places (Whyte, 1980). His work, The Social Life of Small Urban Spaces, remains one of the most cited studies on this topic. Whyte started the street life project with colleagues to examine the use of spaces we use so often. Using time lapse photography to capture human and spatial behavior exhibited within the plaza space, their group found common trends between behavior and the elements on site. Studying these spaces of social engagement is important to developing design standards in order to keep these spaces vibrant. Anderson’s (2016) study suggested that improvements to the public realm have a significant effect on 7 increasing the amount of users, duration of their stay, and activities observed. Urban Plazas have always been spaces for cultural and social celebrations; because of this, they carry a large importance and significance in the environment (Montgomery, 1998; Whyte 1980; Marcus, 1976). Plazas and public squares have been used throughout time to serve many different roles in the community. Even with increasing urbanization of our environments “it is still hard to overlook the enduring importance of plazas to ritual and social life worldwide” (Cobb, 2016). Before the seventeenth century green spaces were used to graze horse, sheep, and cattle, as well as weekend market places (Cobb, 2016). As societies have evolved so did the use of plazas, which carried more social importance, found in seventeenth century France that used walking in these spaces as a way to see and be seen (Costello, 2007). Now in modern societies, vibrant public plazas are characterized as frequently used spaces for social and political engagement, which reflect the success of their surrounding environment. These spaces remain at the heart of communities, which is a reason why they continue to poses such a great importance to the social sphere of society (Orum, 2009). Plazas in a campus act as a parallel public space to plazas in an urban downtown, providing accessible areas to socialize. Like the urban environment, campus design has dramatically changed over the years to provide students a place to study as well as spaces to escape the world of academia (Gumprecht, 2007). Studies have found that nature seen through the window or experienced outside are restorative. These public spaces for students are a way to combat the mental fatigue found when direct attention has been applied for long periods of time (Felston. 2008; Siu Yu lau. 2014). An example for the importance a well-balanced college campus holds, is shown in Hajrasouliha’ (2016) study, which looked into 103 campus university 8 master plans and the relation to graduation rates. The findings found a correlation with those that scored high in the amount of greenery, urban feel, and livability also scored the highest in the availability of natural environments that provide physical and mental restorative benefits for students. In terms of campus planning this holds valuable information for the order which we rank elements. As design continues to improve the campus structure for students, the plaza will continue to hold one of the highest rankings for future planning. 2.2 Urban Plaza Design Elements Urban Plazas have and will always be changing to fit the needs of current users and the surrounding environment. Therefore the design of these spaces will always be changing to fit these adapting challenges (Siu, 2015). However, the main principle remains the same and that is a public space not only needs to attract people into the space but also needs to also keep them there with entertainment and design elements (Zang, 2010). Effective design also relies on planning around the site to consider such things as sun, wind, and entrance locations in order to provide a space that is protected but also remains open (Whyte, 1980; Ghavampour, 2016). Once these design elements are placed and the ground planes are laid, now the space becomes a comprehensive environment (Ye, 2013). The use of design elements is what turns a space into a destination. Many designers and planners strive to understand which objects to select and how these elements should be arranged in space. For example the use of seating, food vendors, and sculpture drew more people into a space than spaces without these elements (Whyte, 1980; Abdulkarim & Nasar, 2014). While the inadequacies posed by the presence of shelter and shade, arrangement of seating, seating environment, and layout hindered people’s perception of the “publicness” of the public park (Bandara, 2013). In their study, an image of a plaza was manipulated by Photoshop to contain 9 these different elements in combination or separate from one another. The images were shown to random individuals to gather ratings on the “visitibility” for each created. Seating was specifically found to be one of the most important design element a space can have (Whyte, 1980; Adbulkarim, 2014; Anderson, 2016; Ewing, 2016; Ghavampour, 2016). It does not matter how densely populated or sparsely occupied, urban furniture are “indispensable fixtures in cities” (Satir, 2005). A type of seating that has not seen a lot of research but has been one of the most used objects, is the moveable chair. Moveable chairs, also referred to as nomadic furniture, are objects which are free to be moved to fit the needs of the user. Very few researchers explored the use of nomadic furniture in their work (Whyte, 1980). One recent study explored the use of moveable furniture and proposed design strategies to advance this topic was, Wong, (2015). The author found little consideration has been given to this type of furniture even though the needs of society are rapidly changing. Their case study in Hong Kong found that permanent furniture locations could easily cause safety or management problems due to the resistance of change. To combat this issue, their findings propose nomadic furniture as a way to keep up with these changing needs for different societies and cultures. 2.3 Human Behavior Not only do societies undergo change but behavior changes even more suddenly due to environmental and other social factors. Even though behavior is always changing, it changes in notable patterns. In social situations it can be expected that people will deal with a certain level of crowding if it is a space that is preferred. However, when that limit reaches capacity it is known that people will move to areas that are less desirable (Zacharius, 2004). 10 Patterns like this are seen with environmental factors, most recognized is the need for shade when it is hot and the ability to sit in the sun when is chillier (Ye, 2013; Whyte, 1980; Peters, 2010). This pattern was also discovered by Lin (2009) in a study in Taiwan, in which 90% of people visiting the public square in the summer opted to sit under shade trees or building shelters. More study’s’ have been conducted to further highlight behavioral shifts found in response to weather. In fact human behavioral responses can be used as a measure for the satisfaction and sensitivity to weather conditions (de Freitas, 2015). They go even further, stating that it is the effective use of shade and clothing that are the best indicators of heat and cold stress (de Freitas, 2015). Ensuring the comfort of visitors is a large portion of designing in the planning professions. Addressing environmental comfort, Giuffrida (2018) states “that thermal comfort is linked to the psychological responses of people to environmental variables.” in response to Gagge (1967) definition of thermal comfort. Indicators for the presence of people in open spaces were found to be temperature, wind speed, and sky clearness (Thorsson, 2007). Further research has been done by de Motingny et al. (2011) on the effect temperatures, precipitation intensity and sunlight hours had on pedestrian fluctuations in 9 different cities located in the northern hemisphere. These indicators were very important when thinking about seating design because they impact human comfort. Therefore designing stationary seating in a way that allows people to have choice is very important. Additionally, providing the opportunity to respond to the weather in a personalized fashion, with moveable seating is equally important. Therefore, does behavior in urban plazas change when the ability to manipulate furniture is available? Studies have shown that freedom and control are by far the most significant predictors of life satisfaction (Verme, 2009). The ability to choose, has been defined in a perimeter of fixed 11 preference patterns when confronted with a set of alternatives (Arrow, 1977). The more alternatives that are presented, the larger your freedom of choice becomes. Would including plaza elements such as nomadic furniture, which has a high freedom of choice increase people’s happiness in public plazas? Behavior based studies have recorded variables pertaining to social, work, and home lives, however there is limited information on how people behave when freedom of choice is introduced in public spaces. 2.4 Mapping To find out questions like the one previously noted, researchers such as Whyte and Adbulkarim, mapped and collected information pertaining to spatial analysis and behavioral decisions. Human behavior could now be mapped using a variety of techniques that allow the researcher to observe and map the location of users at site. A method that has been used since William Whyte and throughout current research is time lapse photography (Whyte, 1980). This method includes taking photographs in even intervals over time until it forms a stream of events. With the introduction of more innovative technology which offers a bias free way to address multiple characteristic and their relationship to one another, it has changed the way we gather and review data (Adbulkarim, 2014). GIS and GPS are popular new ways of gathering data that allow researchers to track the movements and paths of users while also keeping track of user characteristics (Ye, 2013; Ghavampour, 2016). The ability to map and analyze the use of space is imperative in understanding how to effectively design urban plazas. Using elements such as nomadic furniture may be ways to keep up with changing societies without having to constantly redesign to deal with new challenges. More research has to be done to investigate the use of these design elements to make conclusions about their uses in the coming future. 12 2.5 Design Response to Climate Instability What will come in terms of climate change and inevitably social change in the future will have a big impact on how public spaces are designed. These changes will come for every city, therefor as designers it is imperative to update existing infrastructure and planning to form resilient design (Cerra, 2016). The types of resilient design being discussed in current research revolve around ecology. Mainly storm water management and resilient planting plans (Alizedah, 2019; Koslowski, 2016). However the other sphere that uses these spaces are human users, which will also have to adapt to changing environmental conditions (Cerra, 2016). Much like the new ideologies around water and ecological systems, there is a need to review how design elements need to adapt to keep providing recourses to the social sphere they cater to. However, there is a lack of research on public furniture and design elements in response to climate change. As mentioned earlier previous research from Sui (2015) has found in Hong Kong that immoveable furniture poses safety and management problems due to its lack of flexibility to change with its surroundings. This is why it is necessary to conduct more studies on these types of elements that play a large role in the social life of public spaces. 13 CHAPTER THREE METHODOLOGY Preamble To begin planning for the future, one must understand the needs of the current society, its wants, and what functions must be designed to satisfy it. In this research nomadic furniture will be looked at as a function to serve the public. To understand how human behavior is effected with the introduction of nomadic furniture, a quantitative approach will be used to uncover relationships. Primary data collection will rely on longitudinal observations as well as regression analysis to draw conclusions. This study takes place on Michigan State University’s campus in the plaza between Wells Hall and the International Center (See Figure 5). The project “Idea Chair” was proposed to incorporate nomadic furniture to provide a space for students where they can manipulate their environment to fit their needs. To measure how these chairs are manipulated by the users, time lapse photography was established for the two months the chairs were present. They captured images of the chairs’ movement throughout the course of a day, every day. The images that were collected were coded to follow the chairs movement in relation to the spatial analysis of the site. The analysis should provide an insight into the behavioral patterns exhibited by the visitor’s use of the moveable furniture. 14 Figure 3. Photo showing Adirondack chairs on site. Source: by Author Figure 4. Photo showing Adirondack chairs. Source: by Author 15 3.1 Furniture “Idea Chair” was a project started by Michigan State University landscape services department at the Peoples Park. The project progressed to include a group from landscapes services as well as faculty from landscape architecture and urban planning departments. The research aspect of this project created an opportunity for masters’ students to study aspects of the chairs to better understand how they are being used in the space. For this research, nomadic furniture was appropriate for the location of study, which was in a college campus environment. The type of furniture used was light weight plastic Adirondack chairs in a multitude of colors. The type of furniture that is presented effects the use of a space (Adbulkarim & Nasar, 2014), due to the different perceptions users have. The goal of the chairs was to allow for an expansive amount of uses such as: relaxing, doing homework, group socializing, lunch breaks, and anything else that can involve sitting outside. To promote the movement of the chairs they were made out of light-weight plastic so that there would not be limitations to who could and could not move the furniture (see Figure 4). Every night the furniture would be taken inside and in the morning arranged in the same locations according to a pre-proposed map. This ensured the starting point every day was the same, so that daily movement of the chairs would be comparable to one another. 16 Figure 5. Image showing site of idea chair project. Source: Google Earth 3.2 Site The parameters of the site were loosely comprised of the river trail sidewalk running along the Red Cedar River, the side walk in front of Wells Hall and the side walk running alongside the International center (see Figure 5). The space in these boundaries, was made up of ten green spaces split up by the sidewalk system. There is a heavy amount of trees near the river with a few selectively around the site creating pockets of shade. To the east of the site is Farm Lane, a major road on campus, and to the South West is Red Cedar Rd leading to the MSU stadium and recreational facilities. Chair Setup On site there was a set number of chairs that would be out on any given day. The total number of chairs that were set up each day was 20. These chairs were set up in specific locations around the site each morning for the duration of my study. At the start of each day all the chairs 17 were visible to the time-lapse camera. As the day went on the chairs took on new positions as students and faculty moved them to their desired locations. As the day went on, the data collection captured the number of chairs on site as they moved around the space. Figure 6. Photo showing Adirondack chairs setup for each morning as a control. Source: Google Earth Spatial analysis The site has been described as an open green space, divided into smaller pieces by the interlacing sidewalk systems that weave through the site. The mosaic of grass that creates the interior of the plaza is bordered by major sidewalks and bike paths. The space closer to Wells Hall incorporates a large hardscape patio that provides fixed covered seating. The diversity in the space creates an opportunity to zone the site according to different characteristics. Research 18 shows that depending on the location of certain elements it can effect where visitors prefer to go (Zacharius, 2004; Adbulkarim & Nasar, 2014) which has influenced the types of zones chosen. The zones that will be established consider shade, proximity to the sidewalk, proximity to buildings, location of major intersections, views, and noise. By analyzing the space, the movement of the chairs can be related to larger variables such as time and the types of social interaction. Zone Tracking the chair count for this research relied on creating locations within this study site. Zones were proposed as ways to segment the space into groups that shared similar descriptive attributes. Characteristics that made up this social sphere were physical and conceptual spaces such as edges, interiors, trees, light posts, intersections, and the spaces that border these elements, as I hypothesize that they all play a significant role in influencing the actions exhibited by visitors. By looking at the space in terms of the separate characteristics, a number of zones emerged. Those were based on characteristics of the site as well as their position within the boundaries. A positioning that became a limiting factor when studying a site with a large amount of tree coverage is the inability to see through the canopies, which will be noted as an obscured part of the site to compensate for the lack of visibility. This limitation was solved by implementing two camera positions, so one was above the canopy and one took images below allowing for cross reference analysis. Out of this critical review, seven zones emerged: sidewalk edges, sidewalk middle, obscured, edge intersection, middle intersection, tree edge, and grass. The result was formed by adding a “middle” and “edge” description onto each of the major groups. An example of this would be for “obscured”, where there was also a “middle obscured” 19 and an “edge obscured”. However with certain zones, the space was split into so many sections that the density of chairs became almost nonexistent throughout the study. These zones were merged them together in order for the specific zone to cover a larger area and increase its chair density throughout the study. The visual coding of the zones on site is demonstrated in Figure 7. Figure 7. Visualization of the seven zone group locations. Source: by Author Acreage As previously discussed, zones captured the descriptive characteristics of the site and grouped them into the appropriate categories. Acreage then accounts for the amount of space these design characteristic hold on the research site. To calculate density, acreage was found for a specific zone and divided by the number of chairs that were counted for the specific photo (Acreage / Number of chairs = Density of zone). This was repeated for each zone in each photo. Acreage was collected by using Google Earth and using the polygon tool to obtain measurements. This Software was used due to the angle the photos were recorded at, this was the best way to replicate the vantage point. To increase the accuracy of the measurements both images, google earth and time-lapse photos, were loaded side by side. The measurement for the entire space ensured that the individual measurements neither fell below or above the overall 20 acreage of the site. Tracing the shape of the zones on to the Google Earth maps, this was repeated for each of the seven zones on site. The entire site equaled: 0.60 acers and the acreage for the zones are as follows: 1 Grass: .16 Acres 2 Obscured: .23 Acres 3 Trees Middle: .075 Acres 4 Side walk Middle: .045 Acres 5 Sidewalk Edge: .05 Acres 6 7 Intersection Edge: .03 Acres Intersection Middle: .01 Acres Total: 0.60 Acres Acreage Grass Sidewalk Middle Intersection Middle Obscured Sidewalk Edge Trees Middle Intersection Edge Figure 8. Distribution of land between the seven zones. 21 Density of chairs The chair count was measured by looking at individual photos throughout the day and through utilizing Photoshop the chairs were located on site and counted for each zone. However, the size of each zone differs significantly which may influence the outcome of the chairs represented in individual zones. So, the number of zones was standardized by the size of the zone to account for the fact that some zones covered larger areas than others. This ensured that the results are not purely based on chair count, but also compare the amount of land a certain zone occupies, density was added as an additional variable. This variable measures the amount of chairs per square 100 meters. The next section shows how the acreage of each zone was measured. The acreage was converted into square meters and this variable density was calculated in Stata by dividing the number of chairs in a certain zone by the specific square meters that zone 3.3 Independent Variables Temperature The local temperature for East Lansing was measured every hour for each day the study was conducted. This data was obtained through a historical weather data website, called “Time and Date Weather”1 where they have recorded, the average temperature for the area by the hour, taking into account wind, the amount of sun, rain, and snow. To verify the accuracy of this website, additional weather sites were consulted to compare. However, the other sites only had daily historical temperature data, which could only give a rough representation of its accuracy. The data collected from Time and Date weather was inputted into the excel document containing the time-lapse images and chair count. Hourly data was necessary because this study was looking 1 https://www.timeanddate.com/weather/usa/lansing/historic?month=5&year=2018 22 at randomly selected time-lapse photos, the weather patterns explain a lot for the differences in movement that was observed between these images. This study spanned two months, from the beginning of September through the end of October. Between these months the weather went through significant changes. In early September temperatures remained in the upper 80’s which dropped to low 40 degrees Fahrenheit be the end of the study in October. Within these two months, the morning and evening temperature showed about a 10 degree change throughout the day. The pattern exhibited in the temperature variation will be important in understanding the pattern in chair placement. Figure 9. Temperature range through out the study Weather Weather, specifically the amount of cloud coverage is measured categorically through observational analysis. The categories for this variable included ‘cloudy’, ‘partly cloudy’, and ‘sunny’. These variables were determined by the amount of cloud coverage and the amount of shadows that are created on the ground. The weather was considered cloudy if the majority of the 23 sky was covered with clouds and there were no shadows produced on the ground plane. Weather was categorized as partly cloudy if part of the sky was cloudy still producing shadows on the ground. If the day was described as sunny then there was no clouds in the sky to cause the sun to be covered up. The weather was observed for each photo at the same time the chair data was collected. This is important because the amount of sun or shade can effect chair placement in relation to temperature. Time of day Another variable that had consistent variation was time throughout the day. Time was predominantly captured through time lapse photography. The time lapse camera was set to take photos every 10 minutes, every day from 7:00 am to 9:00 pm throughout the two months of the study. The 14 hour time period produced 3,360 photographs. However through photo analysis the time was condensed down to 9:00 am until 7:00pm. This was due to the limited lighting in the morning and evening hours which created problems locating the chair placement on site. The reasoning also included the limited number of users during both of those time periods, resulting in no chair moving until later in the morning, this reduced the number of photos to 2,400. 3.4 Data Collection Time Lapse photography Inspired by previous research that used time lapse photography to capture human behavior and movement in a space, this research used the same approach to observe the people moving chairs (Whyte. 1980; Ghavampour, 2016). Time lapse photography captures human behavior without approaching the subject which helps limit the amount of subjectivity. Due to the limited human contact, an IRB form was submitted and reviewed as exempt. The time-lapse 24 cameras were BRINNO TLC200 Pro models and were set up accordingly: Time lapse frame rate: 30 FPS, White Balance: Auto, Image Quality: Best, Scene: Daylight, Timer: None, HDR Range: High (Resolution), Exposure: Left in the middle, ASAP: 10 min. As discussed previously two time lapse cameras were employed to capture a view from a bird’s eye view and one that captured movement under the tree canopy. This was executed by positioning the cameras in the two buildings on either side of the green space as seen in Figure. 10. One of the cameras was located in an office window on the seventh floor of Wells Hall, this provided the birds-eye view of the space. The second camera was placed in an office windowsill on the second floor of the International center that captured a lower angle that the first camera may have missed The cameras were set to take images every 10 minutes from 7:00 am to 9:00 pm every day. This study went on for two months starting at the beginning of September and ending in the last week of October. The researcher of this study would return to these cameras every three days to download the images and replace the batteries. The time lapse photos collectively make short films comprised of the individual snapshots of the days’ activities, which provide a unique narrative to study in relation to the site. 25 Figure 10. Locations of the two cameras used to capture images above and below the tree canopy. Figure 11. (Left) Typical image captured by BRINNO camera. (Right) Image of the specific camera used in study. Source: www.bhphotovideo.com Random Selection of Data Random selection was employed because a reduction in the amount of photos would still result in a large data set while also allowing efficient analysis. One image was randomly selected from each hour during the study period, reducing the final photograph count was reduced to 500. A discussion later in this chapter will go into more detail on how random selection was 26 administered. As discussed previously in the description of variables used in this study, the original number of photographs was reduced to 500 by random selection. To prepare for the random selection process, the 10 hours of data collection were separated into 10 different groups that contained approximately 1 hour into the variable time group “time groups”. Random selection was administered to each of the time groups, selecting one photo from each group. This stratified random sample separated the day into multiple periods, thus ensuring that when random selection was applied there would be an equal distribution of times picked throughout the day. The results then would be able to create an unbiased picture of what hourly chair use looks like day to day. Thus, one day with 8 time groups (1-8) and 7 zones per time group results in 56 observations per day (1 day = 10 time groups X 7 zones on site = 70 observations). 3.5 Visual Analysis With the final images chosen, coding was the next step in the process that began to look at chair placement in regard to spatial analysis. This process consisted of using multiple layers in Photoshop to recognize where the chairs were in relation to the designated zones. The first step in the process was importing one of the selected photos into Photoshop. Next, a grid was applied to the photograph to first create the zones that would be used for every photograph. The third step required each visible chair to receive a black dot that denoted its position on site. In this research individual chairs will not be tracked on site due to the lack of detail captured and lapse in time that. Coding the dots will allow us to see where the chairs move in relation to the different zones established, hence developing a relationship between the nomadic furniture and the behavior of the visitors. Finally, the zones were applied over the site, but under the dots. This was used to take count of the chairs in each zone to input into the excel document. 27 Figure 12a. Demonstrates step one in coding chairs. Photo taken from time lapse camera. Figure 12b. Demonstrates step two in coding chairs. Grid overlaid on picture to set a base for the application of zones. 28 Figure 12c. Demonstrates step three in coding chairs. Dots are overlaid on top of chairs. Figure 12d. Demonstrates step four in coding chairs. Zones are added. Chairs are tallied for each zone and added to excel sheet. 3.6 Statistical Methods Statistical analysis will be used to address the main research question which is testing if abiotic factors affect behavioral responses on moveable furniture. The following models are 29 testing three main hypotheses: the effect of time, temperature, and weather on user behavior. The hypothesis for these three variables are as follows: Hypothesis 1: Increases in temperature will lead to increases in the density of chairs in shaded zones and decreases in the density of chairs in unshaded zones. Hypothesis 2: The morning, afternoon, and evening time groups will show distinct increases in the chair density of certain zones over others. Hypothesis 3: A decrease in the amount of cloud coverage will lead to an increase in the density of chairs in shaded zones and a decrease in density of chairs in more exposed zones. To test these hypothesis two different models were used. Preliminary results and patterns were examined through descriptive statistics and scatter plots. The patterns that showed up in descriptive statistics were analyzed further through regression models with interaction terms between key independent variables (time, temperature, and weather) and the type of zone. These allowed the study to quantitatively describe the relationships seen between zones, abiotic factors, and the placement of chairs. The equation for the interaction models is described below. Regression Models Ordinary Least Squares regression analysis was used to examine whether the patterns observed in the scatter plots illustrated statistically significant patterns of chair placement; that is, whether the results allowed me to draw conclusion about the placement of chairs during all time periods throughout the study. The full results of these analyses are shown in Tables 3, 4, 5, and 6; however, due to the complicated nature of the interaction regression equation I refer to the 30 graphical representation. Unlike a linear regression with one variable, these models include two variables that interact. Thus producing an equation that reads: Y = B0 + B1X1 + B2X2 + B3X1X2 + E where Y is the dependent variable, chair density. This is the variable that is influenced by either temperature, weather, or time group. B3 represents the Y intercept for the reference category (one of the zones). X1 represents a vector of categorical variables indicating each zone and B1 represents the estimated differences in the Y intercept for each zone (when compared with the reference category). X2 is one of three independent variables (time, temperature, or weather) that is allowed to interact with zones (X1). B2 thus represents the effect of temperature, weather, or time in the reference category (one of the zones). B3 in this equation is the regression coefficient where as X1X2 is the interaction between temperature, time, or weather and zones. If there was no interaction term then B1 would be interpreted as a unique effect of temperature, time, or weather on chair density for all zones. However, the interaction means that the effect of temperature, time, and weather on chair density is different for different zones. So the unique effect of temperature, time, or weather is not limited to B1 but dependent upon the interaction of abiotic factors (X2) and zones (X1), as estimated by B3. Lastly the error term is the difference in what is predicted and what is actually observed. This equation is what will produce the following interaction graphs. Elaborating on the general description of data selection from the above section, a detailed explanation of the technical statistical methods that followed the data collection is presented in the following section. 31 Stata was used to conduct regression analysis examining the hypothesis examining the differential effect of time, temperature, and weather on chair placement in different zones. Starting with the completed excel sheet containing the previously discussed variables (see Figure 13), it was input into the software. However, further cleaning and coding of the data was needed to properly function within this program. Figure 13. Format of excel document before STATA The variables such as temperature and time work within the context of this statistical software, however categorical variables needed to be recoded. A variable that this applied to was Zones; this needed to be recoded so it would be recognized by the program as having numerical labeling for each zone. The zones were thus numbered accordingly going down the list until each zone had a number ranging from 1-10. Later in the analysis process, certain zones were merged together due to number of chairs showing no significant findings. Numerical recoding was also necessary for the weather patterns that were recorded. Originally the data for weather included descriptive labeling such as Cloudy: Yes/No and Sunny: Yes/No. As previously discussed weather would either be cloudy, sunny, or when partly cloudy it was rendered both as Yes. These values became 0 for cloudy, 1 for partly cloudy, and a 2 for sunny. Again this was done in order to utilize these variables as categorical predictors in the regression analysis. The excel document produced, through the addition and subtraction of certain variables, is provided in Figure 20. 32 One of the last steps in coding, took into account that some zones started with more chairs than others. Failing to account for the number of chairs in each zone in the starting time period could bias the estimates of the effect of the independent variable. This can be the case because of the fixed placement of chairs in the morning setup. Such that, more chairs were brought to the grass rather than under trees, not because of user preference but the study’s actions to create a baseline. To account for this difference, code was written into Stata that acknowledged the number of chairs in each zone at time group 1. Since the chairs were set up in the same position every morning, time group 1 could be used as a consistent source of chair count in the beginning of the day. The code then created another variable that took time group 1 chair counts for each zone and distributed it down the data set so that when running any equation with said code, it would take in account the starting number of chairs so that it rules out the possibility that starting number of chairs has any effect on the final findings. Figure 14. Shows output from STATA For the variables and times that had no chair count added, due to either rain, lighting, camera was moved, or chairs were not put out, it was coded as missing. This missing data is recognized by Stata but cannot be used in a regression model. This is important because if the missing data is not coded in this way, then the results will not be accurate because it would be 33 taking into account numerous times when it looks as though the chair count it zero, when it really should not be registering anything at all. Table 1. Summary of missing data Rain Camera was out of position No No Chairs No Photos To Dark Yes Total Freq. Percentage Cum. 216 3,016 120 120 8 80 2,25 3,560 6.07 84.72 3.37 3.37 0.22 100 6.07 90.79 94.16 97.53 97.75 100 Ensuring the data set is clean is essential to running models that will result in valid findings. The preliminary statistic models run in this research were looking at descriptive summaries of all the variables in question before running inferential statistics. Inferential statistics allowed this study to infer results for the population of students that visit this location on Michigan State University’s campus during the study period; therefore, in regards to external validity, this study can generalize to the pattern of chair usage in this location during the study period. 34 CHAPTER FOUR FINDINGS AND DISCUSSION This study set out to find if abiotic factors had an effect on user behavior as they use moveable furniture. Findings from the statistical models run reveal that there are significant relationships that exist between chair placement, temperature, time, and weather patterns. These findings supported my original hypothesis for these three independent variables. Specifically when temperature was high more chairs were located in the shade, while in lower temperatures more chairs were placed in exposed areas on site. Time of day revealed three distinct time periods when chair movement correlated to a specific zone. When testing weather, the more cloud coverage present the more chairs were out on the open lawn, while the fewer clouds resulted in more chairs moving under the tree canopy. 4.1 Zone Summaries One of the strongest preliminary findings in this study showed up in the popularity seen across the different zones. Mapping the chair placement throughout the complete duration of the study, the majority of chairs resided in ‘near trees’ zones and ‘edge sidewalks’ zones (see Table 2). These results, showing significant evidence for behavioral preference, give reason to start questioning if there is a larger pattern that influences chair placement. 35 Table 2. Summary of chair density per zone Zone Edge Intersection Near Trees Edge Sidewalks Grass Middle Intersection Middle Near Trees Obscured Middle Sidewalks Total Mean 0.0300448 1.128873 1.094623 0.7476946 0 . 0.5752552 0.4089211 0.568781 N 470 470 470 460 470 0 460 770 3270 4.2 Initial Trends The trends that were captured in the initial variable summaries, prompted the examination of temperature, time group, and weather further. Scatter plots were run as descriptive statistical methods looking closer at the reasons in each that influenced chair movement. Running scatter plots for the relationship between chair placement and temperature, evident patterns begin to emerge. Yet again a striking inverse relationship is depicted between ‘grass’ and ‘obscured’ zones (see Figure 15). Within this relationship, a higher number of chair counts are in the ‘grass’ when temperatures drop to around 50 degree Fahrenheit. As the temperature grows warmer the inverse takes shape, with the chairs moving out ‘grass’ and a larger amount of them collect in ‘obscured’. Another pattern emerges when the temperature is not at its extremes but lies more in the middle, ‘edge sidewalks’ and ‘Middle near trees’ pick up more chairs. However they taper off as the temperature moves to the extremes again. 36 Figure 15. Scatter plots depicting the relationship between temperature, zones, and chair count In the scatter plot examining the influence time group has on chair placement, most of the zones remain relatively constant throughout the day. However, slight variations do present themselves. In the morning, ‘edge sidewalk’ and ‘grass’ start off at the highest point for these two zones. Later in the day they go through subtle changes, but the morning remains the peak time for chairs to be placed in these zones. Afternoon created a peak time for chairs to be placed in ‘near trees’ and ‘middle sidewalks’ zones. While in the evening, ‘obscured’ saw a rise in the amount of chairs that got moved there and stayed there for the rest of the day. The two zones that time did not have a significant effect on are “middle intersection’ and ‘edge intersection’. However slight, the variations in seating placement based upon behavior in consideration of time produces noticeable patterns of placement choice. 37 Figure 16. Scatter plot depicting the relationship between time group, zones, and chair count Within this last scatter plot (Figure 17), descriptive statistics were run on the variable ‘weather’ to understand how this environmental condition impacts seating preferences. The weather was coded as 0 if the sky was predominantly cloudy, if the sky was observed to be partly cloudy it was coded as a 1, and if the sky was clear and sunny then it was coded as a 2. Across all the zones, there is very little change concerning the movement in chairs in response to cloudy, partly cloudy, or sunny weather conditions. Other than the outlier in ‘edge intersections’, the most noticeable change is within zone ‘middle sidewalks’. This zone is at its lowest chair count when the weather is noted as cloudy. As the day gets increasingly sunny the chairs start to form a positive trend in chair count. This seems to be the only trend in the group of zones, which preliminary starts to show that weather may not have a significant impact on seating preference. 38 However, further research is needed to see is there is any difference in the interaction of zones in relation to weather. Figure 17. Scatter plot depicting the relationship between weather, zones, and chair count 4.3 Interaction Models The interactions models were used to test the interaction of independent variables to see their effect together on chair density. The following models were testing temperature, time, and weather with zones. B1X1 is representing either temperature, time, or weather and B2 is the effect of these variables when zone equals 0. And the interaction term in the equations is X1X2, which is important to test the effect of temperature, time, and weather on chair density which will be different for different zones. So the unique effect of temperature, time, or weather is not limited to B1 but dependent upon the interaction of B3 and zones. 39 Interaction model – Temperature Observing the zones all together allows the differences to be highlighted. Here, three zones show significant effects brought on by change in temperature. These include ‘grass’, ‘edge sidewalks’, and ‘obscured’. The other four zones including ‘edge intersections’, ‘middle intersections’, ‘middle sidewalks’, and ‘edge near trees’ show little response to changes in temperature. For these three zones discussed in this section, the y-intercept is when the temperature is 35 F0. In the case of ‘grass’ and ‘edge sidewalks’ their relationship seems to be very close in nature, they start at a 0.5 chair/m2 difference in terms of their y intercept when the temperature is 35 F0. At this point these zones have an average of 1.5 to 2.00 chairs/ m2, which decreases to about .25 chairs/ m2 when temperatures reaches 90 F0. This model illustrates for both of these zones that when temperature goes up 20 F0 these zones on average loose about .5 chairs/ m2. The inverse to this pattern is noted in the zone ‘obscured’ where the chair count/ m2 is around 0 when the temperature is 35 F0 (again the y-intercept in this model), but gains a considerable amount of chairs when temperatures reach 90 F0 to around 1.5 chairs/ m2. The average rate that this zone loses chairs is around .5 chairs/ m2 lost per 20 F0 it goes down. Within this model we are 95% confident that the density of chairs will remain in this range for each zone represented in the model. The pattern repeated in the majority of the zones shows a wider confidence interval closer to the extreme ranges of temperature, while there is a narrower interval closer to the middle temperatures on each of the zones. This can be linked to the more extreme temperatures, where more varied choice was observed, while in more moderate temperatures there was more of a majority preference in seating choice. This pattern can possibly also be attributed to the sample size in different temperature ranges, such as few observations at the extremes of temperature. The linear regression for this interaction model shows this model is 40 satisfactory with a P value of 0.000. This indicated that we would see a 0.000 chance that these findings would occur if there was no relationship between temperature, zone, and chair density. Figure 18. Interaction model for Temperature, zones, and chair count Table 3. Linear regression for Temperature, zone, and chair count Source Model Residual Total df MS = = 14 68.3362575 = 0.30890754 0.66897669 4 = = 3 = SS 956.70 761 812.73 575 1769.4 434 2,631 2,645 41 2,646 221.22 0 0.5407 0.5382 0.55579 Table 3 (cont’d) ChairsPer100SqMnew ZoneNew Near Trees Edge Sidewalks Grass Middle Intersections Obscured Middle Sidewalks temperature ZoneNew#c.temperature Near Trees Edge Sidewalks Grass Middle Intersections Obscured Middle Sidewalks Coef. 1.28186 2.99596 Std. Err. 0.22844 06 0.23498 45 t P>t 5.61 12.75 0 0 0 2.29757 0.2297 10 0.07882 6 - 0.89156 94 0.69028 56 0.00175 4 - 0.00343 73 - 0.03246 1 - 0.02621 78 - 0.00175 4 0.02223 73 - 0.00501 86 0.22841 16 0.22844 13 0.22873 17 0.00239 98 0.00340 21 0.00341 59 0.00339 47 0.00339 39 0.00339 4 0.00339 48 0.35 0.73 -3.9 3.02 0 0.00 3 0.73 0.46 5 0.31 2 -1.01 -9.5 -7.72 -0.52 6.55 -1.48 0 0 0.60 5 0 0.13 9 [95% Conf. 0.8339 19 2.5351 87 1.8471 59 - 0.3690 586 - 1.3395 12 0.2417 734 - 0.0029 518 - 0.0101 084 - 0.0391 592 - 0.0328 744 - 0.0084 09 0.0155 822 - 0.0116 753 Interv al] 1.7298 02 3.4567 33 2.7479 81 0.5267 106 - 0.4436 267 1.1387 98 0.0064 598 0.0032 338 - 0.0257 629 - 0.0195 611 0.0049 01 0.0288 924 0.0016 381 42 Table 3 (cont’d) StartingChairPer100SqM _cons Interaction model – Time Group 0.25471 07 - 0.07882 6 0.01575 17 0.16151 14 16.17 0 0.62 6 -0.49 0.2238 237 - 0.3955 282 0.2855 977 0.2378 762 Unlike the interaction model for temperature, we see distinct variation presenting its self with in 5 of the seven time groups. These zones include ‘grass’, ‘obscured’, ‘middle sidewalks’, ‘edge sidewalks’, and ‘near trees’. However, similar to temperature two similar zones, ‘edge intersection’ and ‘middle intersection’ show they are not significantly affected by the change in time throughout the day. The variables ‘edge sidewalks’ and ‘grass’ have a similar difference of .25 chair count/m2 to begin with at the y-intercept, when time is 1 and the zones are 0 for each individual line. Both zones begin with 1-1.25 chairs/ m2 when the day begins. For both these zones, they go through a steep transition in the amount of chairs placed within their boundaries from time group 1-4, where they lose around .25-.5 chairs/ m2 within this 4 hour time frame. After time group 4 the rate at which they decrease begins to become less steep, now only losing about .25 chairs/ m2 over a 5.5 hour time. At the end of the day in time group 8, both these zones stay at a little less than .5 chairs/ m2 per zone. Obscured on the other had starts slightly below .5 chairs/ m2 in the beginning of the day, reaching to just over .75 chair/ m2 by the end of the day. This zone experiences an increase in .25 chairs/ m2 for the first 4 hours the day goes on. Then evens out by the end of the day remaining consistent in number of chairs with possibly only a slight variation. The zone ‘middle near trees’ has a very distinct pattern in user behavior. While this zone starts slightly below 1 chair in time zone one, it only drops around .10 chairs/ m2 by the end of the day in time zone 8. The dramatic change occurs in the middle of the day between time 43 zone 3 and 6, where chair count reaches 1.2 chairs/m2 on average. In this zone the chair count rises and falls on average .25 chairs/ m2 in one hours’ time window. The last zone to show significant difference in chair count according to the time of day is ‘middle sidewalks’. This zone begins the day with about .5 chairs/ m2 and ends the day with roughly the same. However, the trend happens between time group 4 and 7 where this zone sees a decrease in the amount of chairs/ m2 by 0.15. Before and after these time groups the chair rises back to .5 chairs/ m2. Unlike the last interaction model, the 95% confidence intervals show for each zone do not exhibit a recognizable pattern. Rather they remain quit large throughout the whole day for each zone. This explains that throughout the day the chairs did not consistently get placed in a certain zone during a specific hour, instead there was quite a bit of variating in the preferences throughout the day. As in the regression for temperature, time group has a P value of 0.000 which illustrates again that these findings have a 0% chance of occurring if there was no relationship between time of day, zone, and chair density. 44 Figure 19. Interaction model for time group, zones, and chair count Table 4. Linear regression for time group, zones, and chair count Source Model Residual Total SS 917.49968 1101.4219 2018.9216 df 70 3,199 3,269 MS = = 13.1071383 = 0.344301942 = = 0.617596081 = 3,270 38.07 0 0.4545 0.4425 0.58677 ChairsPer100SqMnew ZoneNew Near Trees Coef. 0.700837 Std. Err. 1 0.121997 t P>t 5.74 0.000 [95% Conf. 0.461636 8 Interval] 0.940037 3 45 Table 4 (cont’d) Edge Sidewalks Grass Middle Intersections Obscured Middle Sidewalks timegroup ZoneNew#timegroup 8.30 0.000 5.60 0.000 -0.00 1.000 1.73 0.084 2.16 0.031 -0.00 1.000 0.14 0.885 0.58 0.563 0.58 0.563 0.14 0.885 0.14 0.885 0.89 0.374 -0.00 1.000 -0.00 1.000 0.780205 4 0.446242 8 - 0.237327 6 - 0.028310 9 0.023758 2 - 0.237327 6 - 0.219802 4 - 0.167226 8 - 0.167226 8 - 0.219802 4 - 0.219802 4 - 0.129656 9 - 0.237327 6 - 0.237327 6 1.262783 0.9270472 0.2373276 0.4492541 0.498936 0.2373276 0.2548528 0.3074284 0.3074284 0.2548528 0.2548528 0.3449983 0.2373276 0.2373276 1.021494 0.123062 1 0.686645 0.12261 -9.56E- 14 0.121041 9 0.210471 6 0.121784 0.121175 2 0.261347 1 -7.52E- 14 0.121041 9 0.017525 2 0.121041 9 0.070100 8 0.121041 9 0.070100 8 0.121041 9 0.017525 2 0.121041 9 0.017525 2 0.121041 9 0.107670 7 0.121041 9 -7.57E- 14 0.121041 9 2 3 4 5 6 7 8 9 -7.61E- 14 0.121041 9 10 46 Table 4 (cont’d) Near Trees# 2 Near Trees# 3 Near Trees# 4 Near Trees# 5 Near Trees# 6 Near Trees# 7 Near Trees# 8 Near Trees# 9 Near Trees#10 Edge Sidewalks# 2 Edge Sidewalks# 3 Edge Sidewalks# 4 Edge Sidewalks# 5 Edge Sidewalks# 6 Edge Sidewalks# 7 - 0.298077 9 - 0.157876 4 - 0.187919 6 - 0.187919 6 - 0.225473 6 - 0.473329 8 - 0.555964 6 - 0.410739 9 - 0.500869 4 - 0.283056 3 - 0.563459 4 - 0.910458 1 0.22 0.826 1.04 0.299 0.86 0.388 0.86 0.388 0.64 0.520 -0.80 0.421 -1.29 0.198 -0.44 0.661 -0.97 0.334 0.31 0.759 -1.33 0.183 -3.36 0.001 -3.85 0.000 - 0.994579 -4.10 0.000 - 1.036639 -4.65 0.000 - 1.131275 0.3731859 0.5133874 0.4833442 0.4833442 0.4457902 0.197934 0.1152993 0.2605239 0.1703944 0.3882075 0.1078045 -0.2391943 -0.3233152 -0.3653756 -0.4600117 0.037554 0.171179 1 0.177755 5 0.171179 1 0.147712 3 0.171179 1 0.147712 3 0.171179 1 0.110158 3 - 0.137697 9 - 0.220332 6 - 0.075108 - 0.165237 5 0.052575 6 - 0.227827 4 - 0.574826 2 - 0.658947 1 - 0.701007 6 - 0.795643 6 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 47 Table 4 (cont’d) Edge Sidewalks# 8 Edge Sidewalks# 9 Edge Sidewalks#10 Grass# 2 Grass# 3 Grass# 4 Grass# 5 Grass# 6 Grass# 7 Grass# 8 Grass# 9 Grass#10 Middle Intersections# 2 Middle Intersections# 3 Middle Intersections# 4 -5.17 0.000 - 1.221421 -0.5501572 -4.67 0.000 -4.36 0.000 0.31 0.755 -0.49 0.623 -1.93 0.054 -2.51 0.012 -2.81 0.005 -3.26 0.001 -3.71 0.000 -3.20 0.001 -3.12 0.002 0.00 1.000 -0.10 0.918 -0.41 0.682 - 1.134781 - 1.082205 - 0.283732 6 - 0.422124 4 - 0.669429 6 - 0.770151 8 - 0.821655 9 - 0.898876 2 - 0.975592 1 - 0.888065 8 - 0.874636 2 - 0.335631 9 - 0.353157 1 - 0.405732 7 -0.4635167 -0.4109411 0.3911696 0.2527777 0.0054725 -0.0952497 -0.1467537 -0.2239741 -0.30069 -0.2131637 -0.1997341 0.3356319 0.3181067 0.2655311 - 0.885789 1 - 0.799148 6 - 0.746573 0.053718 5 - 0.084673 3 - 0.331978 5 - 0.432700 7 - 0.484204 8 - 0.561425 2 - 0.638141 - 0.550614 8 - 0.537185 1 9.73E-14 - 0.017525 2 - 0.070100 8 0.171179 1 0.171179 1 0.171179 1 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.171179 1 0.171179 1 0.171179 1 48 Table 4 (cont’d) Middle Intersections# 5 Middle Intersections# 6 Middle Intersections# 7 Middle Intersections# 8 - 0.070100 8 - 0.017525 2 - 0.017525 2 - 0.107670 7 Middle Intersections# 9 9.79E-14 Middle Intersections#10 9.83E-14 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 Obscured# 2 Obscured# 3 Obscured# 4 Obscured# 5 Obscured# 6 Obscured# 7 Obscured# 8 Obscured# 9 Obscured#10 Middle Sidewalks# 2 Middle Sidewalks# 3 0.002335 6 0.172106 9 0.071227 1 0.172106 9 0.149444 5 0.259217 1 0.337484 1 0.363175 6 0.240331 8 0.392378 7 0.343331 4 - 0.070100 8 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.172106 9 0.171179 1 0.005841 7 0.171179 1 49 -0.41 0.682 -0.10 0.918 -0.10 0.918 -0.63 0.529 0.00 1.000 0.00 1.000 0.01 0.989 0.41 0.679 - 0.405732 7 - 0.353157 1 - 0.353157 1 - 0.443302 6 - 0.335631 9 - 0.335631 9 - 0.335115 5 - 0.266223 9 - 0.188006 6 - 0.078234 0.87 0.385 1.51 0.132 1.96 0.050 0.000033 0.025724 2.11 5 0.035 - 0.097119 2 0.054927 6 0.005880 3 - 0.405732 7 - 0.329790 2 1.40 0.163 2.28 0.023 1.99 0.046 -0.41 0.682 0.03 0.973 0.2655311 0.3181067 0.3181067 0.2279612 0.3356319 0.3356319 0.3397866 0.4086782 0.4868955 0.5966681 0.6749352 0.7006266 0.5777829 0.7298298 0.6807824 0.2655312 0.3414736 Table 4 (cont’d) Middle Sidewalks# 4 Middle Sidewalks# 5 Middle Sidewalks# 6 Middle Sidewalks# 7 Middle Sidewalks# 8 Middle Sidewalks# 9 Middle Sidewalks#10 StartingChairPer100SqM _cons Interaction model – Weather 0.046733 8 - 0.105151 1 - 0.110992 9 0.017525 2 - 0.095987 2 - 0.046733 8 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.171179 1 0.058417 3 0.171179 1 0.360886 1 0.013894 0.085589 6 7.34E-14 - 0.288898 1 - 0.440783 - 0.446624 8 - 0.318106 7 - 0.431619 1 - 0.382365 7 - 0.277214 6 0.333644 1 - 0.167816 0.27 0.785 -0.61 0.539 -0.65 0.517 0.10 0.918 -0.56 0.575 -0.27 0.785 0.34 0.733 25.97 0.000 0.00 1.000 0.3823657 0.2304808 0.224639 0.3531571 0.2396447 0.2888981 0.3940492 0.388128 0.167816 Similar to the last two models the major reactive zones are ‘grass’, ‘obscured’, ‘near trees’, and ‘edge sidewalks’. The zones that showed minute changes were ‘middle intersection’, ‘edge intersection’, and ‘middle sidewalks’. Both ‘near tree’ and ‘edge sidewalks’ start at 1.4 chairs/ m2 and have the highest chair count to begin with when it comes to weather. These two zones however, decrease at two very different rates. ‘Near trees’ loses around .3 chairs/ m2throughout a transition from completely cloudy to complete sunny. However, this zone experiences a spike in activity when the weather is partly cloudy, specifically a .10 - .20 spike in chairs/ m2. ‘Edge sidewalks on the other hand has a steady progression down in chair count as 50 the weather turns from cloudy to sunny, experiencing a .60 chair decrease. Within the interaction model for weather the consistent inverse relationship between ‘grass’ and ‘obscured’ emerges again. In this model ‘grass’ has around 1 chair/ m2 when it is cloudy decreasing the amount of chairs/ m2 to about .20 as the cloud coverage becomes sunny. Inversely ‘Obscured’ has around .10 chairs/ m2 when the weather is cloudy and gains around .40 chairs/ m2 as the weather is sunny. Both of these zones differ by around .40 chairs/ m2 as the cloud conditions change to either fully cloudy or completely sunny. As in the previous graph, the 95% confidence values do not seem to carry a specific patter, however there is one confidence interval that is larger than the rest. The confidence interval for ‘edge sidewalk’ when it is cloudy is the largest interval in the model, suggesting that the population has a wide range of preference when it comes to this zone in relation to being cloudy. Finally the P value remains at 0.00 assuring yet again that these findings have a 0% chance of occurring again if there was no relationship between weather, zone, and chair density. 51 Figure 20. Interaction model for weather, zones, and chair count Table 5. Linear regression between weather, zones, and chair count Source Model Residual Total SS 898.56668 870.87668 1769.4434 df 21 2,624 2,645 MS = = 42.7888893 = 0.331888978 = = 0.668976693 = ChairsPer100SqMnew ZoneNew Coef. Near Trees Edge Sidewalks Grass 1.159436 1.191326 0.7887206 52 Std. Err. t P>t .074342 15.60 .0797339 14.94 .0739469 10.67 0 0 0 Interval] 2,646 128.93 0 0.5078 0.5039 0.5761 [95% Conf. 1.013661 1.305211 1.034978 0.6437205 1.347673 0.933720 6 Table 5 (cont’d) Middle Intersections Obscured Middle Sidewalks Weather PartlyCloudy Sunny ZoneNew#Weather Near Trees#PartlyCloudy -0.00624 0.3687409 0.4024883 0.0823667 0.0295723 -0.0417514 Near Trees#Sunny Edge Sidewalks#PartlyCloudy -0.2714202 -0.4813032 Edge Sidewalks#Sunny -0.6280926 Grass#PartlyCloudy -0.3938521 Grass#Sunny Middle Intersections#PartlyCloud y Middle Intersections#Sunny -0.3730883 -0.0823667 -0.0295723 Obscured#PartlyCloudy 0.2243011 Obscured#Sunny Middle Sidewalks#PartlyCloudy 0.3687676 -0.0492349 Middle Sidewalks#Sunny -0.0955921 .0709128 - 0.09 .0710094 .0714798 5.19 5.63 .080118 .0676442 1.03 0.44 .1136715 - 0.37 .0956646 - 2.84 .1137038 - 4.23 .096038 - 6.54 .1133044 - 3.48 .0956933 - 3.90 .1133039 - 0.73 .0956633 - 0.31 .1133049 1.98 .0957208 3.85 .1133236 - 0.43 .0957073 - 1.00 0.93 -0.1452906 0.2295009 0.2623258 -0.074734 -0.103069 0 0 0.304 0.662 0.713 -0.2646463 0.005 -0.4590059 0 -0.7042614 0 -0.8164104 0.001 -0.616027 0 -0.5607302 0.132810 6 0.507980 9 0.542650 8 0.239467 5 0.162213 6 0.181143 4 - 0.083834 4 -0.258345 - 0.439774 8 - 0.171677 1 - 0.185446 4 0.467 -0.3045408 0.757 -0.2171554 0.048 0.0021251 0 0.1810716 0.664 -0.2714476 0.318 -0.2832615 0.139807 3 0.158010 8 0.446477 1 0.556463 6 0.172977 8 0.092077 4 53 _cons 4.4 Final Model A final model was performed to test temperature, time, weather, and zones together. The findings suggest that the combination of these variables explain why people choose to site in specific locations more than testing these variables separately. Specifically this model can account for 58% of why people choose certain locations to sit than others. In the other models, temperature describes 54%, time describes 45%, and weather describes 50% of the reasoning behind user behavior and preference when using moveable furniture. These results have a P value of 0, again denoting these results have a 0% chance of occurring if there was no relationship between the variables tested. Table 6. Linear regression between temperature, time, weather, zones, and chair count Source MS SS df Table 5 (cont’d) StartingChairPer100SqM 0.2708217 .0163639 16.55 .0501429 0.12 0 0.2387343 0.901 -0.0920836 0.302909 1 0.104563 7 0.00624 Model Residual Total 1027.75152 91 11.29397 741.691833 2,554 0.290404 1769.44335 2,645 0.668977 ChairsPer100SqMnew Coef. Std. Err. t P>t 54 = = = Number of obs F(91, 2554) Prob > F R- squared = Adj R- squared = Root MSE = [95% Conf. 2,646 38.89 0 0.5808 0.5659 0.53889 Interval] Table 6 (cont’d) ZoneNew Near Trees Edge Sidewalks Grass 1.12611 2.780403 2.14815 0.2367347 4.76 0.2428797 11.45 0.2378582 9.03 0 0 0 Middle Intersections 0.0814022 0.2366747 0.34 Obscured Middle Sidewalks Weather PartlyCloudy Sunny ZoneNew#Weather Near Trees#PartlyCloudy Near Trees#Sunny Edge Sidewalks#PartlyCloudy -0.7208901 0.2367185 -3.05 0.2369611 2.94 0.696914 0.0682517 0.0778829 0.88 0.0146422 0.0681128 0.21 -0.0832715 0.1105943 -0.75 -0.2990086 0.0963748 -3.1 -0.2796165 0.1104464 -2.53 0.731 0.002 0.003 0.381 0.83 0.452 0.002 0.011 Edge Sidewalks#Sunny -0.4944774 0.0964491 -5.13 0 Grass#PartlyCloudy -0.2378762 0.1101446 -2.16 -0.2324199 0.0963444 -2.41 Grass#Sunny Middle Intersections#PartlyCloudy -0.0682517 0.1101431 -0.62 Middle Intersections#Sunny -0.0146422 0.096326 -0.15 Obscured#PartlyCloudy Obscured#Sunny Middle Sidewalks#PartlyCloudy Middle Sidewalks#Sunny timegroup 2 0.1121218 0.1101453 1.02 0.2736275 0.0963792 2.84 -0.024651 0.1101589 -0.22 -0.0386867 0.096357 -0.010121 0.1218563 -0.08 -0.4 0.031 0.016 0.536 0.879 0.309 0.005 0.823 0.688 0.934 55 0.6618983 1.590321 2.304142 3.256665 1.681735 2.614564 - 0.3826917 0.5454961 - -1.18507 0.2567104 0.2322587 1.161569 - 0.0844684 0.2209718 - 0.1189198 0.1482041 -0.300135 0.1335921 - 0.4879892 - 0.4961902 - 0.6836037 - 0.1100279 - 0.0630429 - 0.3053511 - 0.0218944 - 0.0434988 -0.453858 - 0.4213409 - 0.2842305 0.1477272 - 0.2035273 0.1742429 - 0.1038614 0.328105 0.0846381 0.4626169 - 0.2406609 0.1913589 - 0.2276326 0.1502592 - 0.2490682 0.2288263 Table 6 (cont’d) 3 4 5 6 7 8 9 10 ZoneNew#timegroup Near Trees# 2 Near Trees# 3 Near Trees# 4 Near Trees# 5 Near Trees# 6 Near Trees# 7 Near Trees# 8 Near Trees# 9 Near Trees#10 Edge Sidewalks# 2 -0.0001339 0.1223959 0 0.0544738 0.1249051 0.44 0.0515834 0.1264254 0.41 -0.0151597 0.1294523 -0.12 -0.0147186 0.132546 -0.11 0.1126091 0.1316674 0.86 -0.0279226 0.1311285 -0.21 -0.0205576 0.1305264 -0.16 0.0434582 0.172348 0.25 0.1900843 0.1731151 1.1 0.173557 0.1766606 0.98 0.1527911 0.1788185 0.85 0.191208 0.1830908 1.04 -0.0246266 0.1874653 -0.13 -0.1281511 0.1862416 -0.69 0.047634 0.1854828 0.26 -0.123985 0.1846144 -0.67 0.999 0.663 0.683 0.907 0.912 0.392 0.831 0.875 0.801 0.272 0.326 0.393 0.296 0.895 0.491 0.797 0.502 0.1191189 0.1723364 0.69 0.49 Edge Sidewalks# 3 -0.1536652 0.1731126 -0.89 Edge Sidewalks# 4 -0.4919861 0.1766963 -2.78 Edge Sidewalks# 5 -0.5865836 0.1788594 -3.28 Edge Sidewalks# 6 -0.5001772 0.1831625 -2.73 Edge Sidewalks# 7 -0.5612358 0.1875669 -2.99 0.375 0.005 0.001 0.006 0.003 56 - 0.2401392 0.2398714 - 0.1904518 0.2993994 - 0.1963233 0.2994901 - 0.2690019 0.2386825 - 0.2746271 0.2451899 - 0.1455766 0.3707949 - 0.2850515 0.2292064 - 0.2765059 0.2353908 - 0.2944978 0.3814142 -0.149376 0.5295446 - 0.1728555 0.5199696 - 0.1978528 0.5034351 - 0.1678136 0.5502295 -0.392226 0.3429728 -0.493351 0.2370488 - 0.3160779 0.4113459 - 0.4859942 0.2380241 - 0.2188144 0.4570522 - 0.4931205 0.18579 - 0.8384687 - 0.9373078 - 0.8593394 -0.141015 - 0.9290346 - 0.1455035 - 0.2358595 - 0.1934371 Table 6 (cont’d) Edge Sidewalks# 8 -0.6756322 0.1863097 -3.63 0 Edge Sidewalks# 9 -0.6146091 0.1855201 -3.31 Edge Sidewalks#10 -0.6356192 0.1846308 -3.44 Grass# 2 Grass# 3 Grass# 4 Grass# 5 Grass# 6 Grass# 7 Grass# 8 Grass# 9 Grass#10 0.1331668 0.1723315 0.77 0.0302845 0.1730959 0.17 -0.1839624 0.1766429 -1.04 -0.2771323 0.1787931 -1.55 -0.2369346 0.183074 -1.29 -0.2847631 0.1874486 -1.52 -0.3917722 0.1862059 -2.1 -0.3127086 0.1854439 -1.69 -0.3481095 0.1845969 -1.89 Middle Intersections# 2 0.010121 0.1723309 0.06 Middle Intersections# 3 0.0001339 0.1730939 0 Middle Intersections# 4 -0.0544738 0.1766425 -0.31 Middle Intersections# 5 -0.0515834 0.1787925 -0.29 Middle Intersections# 6 0.0151597 0.1830732 0.08 Middle Intersections# 7 Middle Intersections# 8 0.0147186 0.1874483 0.08 -0.1126091 0.1862059 -0.6 0.001 0.001 0.44 0.861 0.298 0.121 0.196 0.129 0.035 0.092 0.059 0.953 0.999 0.758 0.773 0.934 0.937 0.545 Middle Intersections# 9 0.0279226 0.1854437 0.15 0.88 Middle Intersections#10 0.0205576 0.1845922 0.11 Obscured# 2 Obscured# 3 -0.0681465 0.1723313 -0.4 -0.0535671 0.1730951 -0.31 0.911 0.693 0.757 57 - 0.3102987 - 0.2508239 - 0.0266423 -1.040966 - 0.9783942 - 0.9976604 -0.273578 - 0.2047569 0.4710904 -0.309138 0.369707 - 0.5303404 0.1624155 - 0.6277264 0.0734618 - 0.5959233 0.122054 - 0.6523299 0.0828036 - 0.7569021 - 0.6763442 0.050927 - 0.7100844 0.0138654 - 0.3278015 0.3480434 - 0.3392848 0.3395526 - 0.4008509 0.2919033 - 0.4021764 0.2990097 - 0.3438273 0.3741468 - 0.3528476 0.3822848 - 0.4777389 0.2525207 - 0.3357127 0.3915578 - 0.3414081 0.3825232 - 0.4060699 0.2697769 - 0.3929881 0.2858539 Table 6 (cont’d) Obscured# 4 Obscured# 5 Obscured# 6 Obscured# 7 Obscured# 8 Obscured# 9 Obscured#10 0.0268367 0.176646 0.15 0.1433688 0.1787947 0.8 0.2284628 0.1830765 1.25 0.3480479 0.187454 1.86 0.1729275 0.1862123 0.93 0.4099125 0.1854486 2.21 0.3987297 0.1845942 2.16 Middle Sidewalks# 2 -0.0607388 0.1723312 -0.35 Middle Sidewalks# 3 0.0369699 0.173094 0.21 Middle Sidewalks# 4 0.1070724 0.1766426 0.61 Middle Sidewalks# 5 -0.0638861 0.1787925 -0.36 Middle Sidewalks# 6 -0.0374778 0.1830735 -0.2 Middle Sidewalks# 7 0.1723228 0.1874484 0.92 Middle Sidewalks# 8 0.0222383 0.186206 0.12 Middle Sidewalks# 9 0.0800238 0.1854442 0.43 Middle Sidewalks#10 temperature ZoneNew#c.temperature Near Trees Edge Sidewalks Grass 0.2070213 0.1845942 1.12 0.0026634 0.48 0.001268 0.0002345 0.0037786 0.06 -0.0190777 0.0037861 -5.04 -0.0189696 0.003767 -5.04 Middle Intersections Obscured -0.001268 0.0037666 -0.34 0.0152363 0.0037666 4.05 Middle Sidewalks -0.0054398 0.0037669 -1.44 58 0.879 0.423 0.212 0.063 0.353 0.027 0.031 0.725 0.831 0.544 0.721 0.838 0.358 0.905 0.666 0.262 0.634 0.951 0 0 0.736 0 0.149 - 0.3195472 0.3732206 - 0.2072284 0.4939661 - 0.1305307 0.5874564 - 0.0195293 0.7156251 -0.192215 0.53807 0.0462676 0.7735574 0.0367601 0.7606992 - 0.3986618 0.2771842 - 0.3024489 0.3763887 - 0.2393048 0.4534496 - 0.4144791 0.286707 - 0.3964653 0.3215097 - 0.1952435 0.5398891 - 0.3428918 0.3873683 - 0.2836125 0.4436601 - 0.1549482 0.5689908 - 0.0039547 0.0064906 - 0.0071749 0.0076439 - 0.0265019 - 0.0263563 - 0.0086539 0.006118 0.0078504 0.0226223 - 0.0128263 0.0019467 - 0.0116536 - 0.0115829 Table 6 (cont’d) StartingChairPer100SqM 0.2554168 0.0155825 16.39 0 _cons -0.0814022 0.1673543 -0.49 0.627 0.2248612 0.2859725 - 0.4095661 0.2467618 59 CHAPTER FIVE CONCLUSION The findings from this study on nomadic furniture through time lapse photography will add a body of knowledge surrounding human behavior and design elements that in the past has been understudied. The selected plaza on Michigan State University’s campus is an ideal platform to experience an active and highly visited site. Urban plazas are a topic rising in popularity in academia due to the increasing number of people moving to cities. Since public plazas are a popular social sphere of the urban environment it is important to find what makes them successful. The findings from this study will add to the body of landscape design knowledge, specifically design furniture. The results will frame the patterns observed when nomadic furniture is added to a space. The patterns that develop will give clues for the more successful placement of furniture from observations of human preferences. This will also provide a successful example for designers looking to apply nomadic furniture to public spaces. The patterns that were observed include relationships between zones and temperature, time of day, and weather. This study provided evidence that there is a significant impact these three variables have on people’s behavior in relation to the spatial programming of the site. Temperature showed a strong relationship between locations on site that were very exposed to the sun or near the edges by the sidewalk and locations under trees in relation to the transition in temperature from 42 F0 to 95 F0. A strong inverse relationship was present between these three zones. The zones that had the maximum exposure on the grass and on the edge near the sidewalk had the most chairs when temperatures were below 65-70 F0, slowly decreasing in chair count as temperatures began to cool off. On the opposite side of the spectrum, spaces 60 predominantly under canopy coverage had very few chairs when temperatures were below 65-70 F0 but gained a significant amount of chairs as temperatures increase. Time groups showed that there were three distinct periods of the day in which particular zones gained chairs. These three time groups consisted of the morning hours between the 1-3 time groups, the afternoon between 3-6, and the evening between 6-8. In the morning, again zones that were predominantly located on the open lawn and near the edge by the sidewalks were the most populated with chairs. Transitioning to the afternoon, edges that were near the sidewalks rose in popularity. While in the evening locations on site under the tree canopy picked up the most chairs by the end of the day. Weather also showed that there were significant patterns that correlated with the type of cloud coverage that was present for the time and day. If the cloud coverage was present to where there is no shadows on the ground surface then on the edges of tree canopies and sidewalks were among the locations with the highest chair count. When the sky begins to clear and becomes partly cloudy, the zones that shows the most dramatic change was the space surrounding the edge of the tree canopy which further increases its chair count. Yet again there is an inverse relationship between spaces located on the grass and under the tree canopy where there is a steady increase and decrease shown between these two zones as cloud coverage changes. When cloud coverage is present, density of chairs on the middle of the lawn had a higher number relatively to those under tree canopy. However, as the conditions change to sunny, the reverse occurs and the density of chairs under tree canopies has higher counts compared to those on the open lawn. When all three independent variables were interacting with zones then 58% of the reasoning behind the difference in chair density is explained. This is significantly important for designers 61 because of the planning implications that follow this finding. If designers can understand at least 58% of the behavioral responses to three factors, temperature, time, and weather in relation to seating preference, then there will at least be a 50% chance of designing a successful plaza. The preferences expressed in this study confirm that chair placement is dependent upon expressed user behavior in response to abiotic factors. There for chair placement is important to user behavior in response to where they prefer to sit in relation to temperature, weather, and time of day. These three variables can be very dependent upon location and use of the surrounding area. However, in a general sense, the needs of users in this study suggest that strategic planning is important for the success of a public space. 5.1 Design Implications In light of the results presented in this chapter, and knowing that the abiotic factors tested explain 58% of the reason behind seating preferences, what design implications can be derived from these findings? How can campus planners use these findings to enhance usability of campus plazas? Although one of the limitations of the study was that it only addressed one plaza on MSU’s campus, essentially only covering one part of campus, my findings still provide evidence regarding a few design suggestions that may address seasonal changes, climate change, design and social cohesion, and tactical urbanism, such as the following: 1. The university may consider the installation of umbrellas throughout the year, that will give the required shade when needed or to protect from rain. 2. Offer advice on planting more trees for the creation of more shaded spaces. 62 3. Begin the conversation to think about providing the spaces that students need in the interiors of buildings especially during the wintertime incorporating some of the findings for outdoor use. 4. Install retractable overhang for the providing summer shade and availability of winter sun. 5.2 Limitations The use of time lapse photography was instrumental in capturing sequential images that allowed for comparison of movement over time. However, due to the positioning, magnification power, and accidental repositioning of the camera, a number of limitations emerged. When the chairs were in full view, one is only able to capture and collect data on the presence of a chair. However, it is difficult to capture gender, activity, and age due to the magnification capability of the camera and also do to the fact that the chairs faced away from the camera at different times of the day. These additional variables could have added descriptive characteristics to the users, which could have further added to the reasoning behind chair placement. Without these factors to consider, it will be possible to understand how the chairs move on a daily basis, but it will not be possible to make a correlation to the characteristics of the users. The location of the camera also posed two limitations in my data collection. Without constant surveillance of the camera, its location within an office put it at risk to be nudged by its occupant or by the nightly cleaning crew. The camera was move out of position on a few occasions, leaving me with only the recordings of a blank wall. If, for the majority of the study, the camera remained in position then the other limitation came with the weather. Rain obscured the 63 window, hence obscuring the view of the camera. However, if there was rain, one could infer that with less people being outside and wanting to use the furniture. So the fact that the window was obscured does not leave me with large gaps in my data. Unless there was a light mist that obscured the window, which resulted in losing out on knowing if people would still use the furniture and how. When the camera was not obstructed or moved, then the other limitation came with the boundaries made for this study site. For the majority of the time the chairs remained within the specific space originally coded for them. The zones this research delineated were only applied to the large grass parcels located closets to Wells Hall. The smaller grass parcels, closer to Farm Lane were left out of the study site due to the size and location away from the buildings and intersection. On a few occasions, a chair or two would be moved to these outside locations. Therefore, it was difficult to include them in the zonal groupings within 4 of the images. During data analysis, another limitation came in the form of set up and take down of the site. In a few cases, there happened to be more chairs than what was originally intended. These occurrences only showed up in the analysis for two days out of the two months during data collection. There were only supposed to be 20 chairs on the study site at a time. However, in the instances when more chairs were found there would be at most 2 or 3 more on site that day. There were also instances when the chairs would be put out late or taken in early. This resulted in 11 images containing zero chairs to analyze. Another limitation was the number of locations in this study. This research only looked at one plaza in a unique setting. Therefore the applied characteristics that come with the site, being a college campus will make it so the application of these findings can only advise design decisions on similar sites. 64 5.3 Future Research Although the regression models performed in this study offer strong evidence that time, temperature, and weather impact users behavior [indicating high internal validity], the findings may suggest the following ideas that can inform future research. First it is worth to note here that the Peoples Park provided easy access to pedestrians and limited access to cars. It is important to suggest that, another study would be needed to see the outcome of nomadic furniture in proximity to street access. Second, future research would be needed to combine the use of time lapse photography with surveys to better understand the demographic who uses the chairs. Third, attach GPS to individual chairs to track real time movement of moveable furniture. 65 BIBLIOGRAPHY 66 BIBLIOGRAPHY Adbulkarim, D & Nasar, J. (2014). Do seats, food vendors, and sculptures improve plaza visitability?. Environment and Behavior, 46(7), 805-825. Doi: 10.1177/001396512475299 Alizadeh, Behdad and James Hitchmough. 2019. "A Review of Urban Landscape Adaptation to the Challenge of Climate Change." International Journal of Climate Change Strategies and Management 11 (2): 178-194. Anderson, J., Ruggeri, K., Steemers, K., & Huppert, F. (2017). Lively social space, well-being activity, and urban design: Findings from a low-cost community-led public space intervention. Environment and Behavior, 49(6), 685-716. doi:http://dx.doi.org.proxy2.cl.msu.edu/10.1177/0013916516659108 Arrow, K. (1977). Current developments in the theory of social choice. Social Research, 44(4), 607-622. Bandara, Viraji. (2013). Perception of “publicness” of the public spaces with special reference to public parks in colombo and sri jayewardenapura. Cities, People Places. Cerra, Joshua Frank. 2016. "Inland Adaptation : Developing a Studio Model for Climate- Adaptive Design as a Framework for Design Practice." Landscape Journal 35 (1): 37-55. Cobb, C. (2016). Mississippian plazas, performance, and portable histories. Journal of Archeological Method and Theory, 24, 676-702. Doi: 10.1007/s10816-016-9281-3 Costello, V. ( 2007). Public spaces for recreation in Dublin, 1660-1760. The Garden History Society, 35(2), 160-179. De Freitas, ,C.R. (2015). Weather and place-based human behavior: Recreational preferences and sensitivity. International Journal of Biometeorology, 59(1), 55-63. doi:http://dx.doi.org.proxy2.cl.msu.edu/10.1007/s00484-014-0824-6 Ewing, R., Hajrasouliha, A., Neckerman, K. M., Purciel-Hill, M., & Greene, W. (2016). Streetscape features related to pedestrian activity. Journal of Planning Education and Research, 36(1), 5-15. doi:http://dx.doi.org.proxy2.cl.msu.edu/10.1177/0739456X15591585 Felston, G. (2008). Where to take a study break on the college campus: An attention to restoration theory perspective. Journal of Environmental Psychology, 160-167. doi:10.1016/j.jenvp.2008.11.006 67 Gagge, A. P., Stolwijk, J. A., & Hardy, J. D. (1967). Comfort and thermal sensations and associated physiological responses at various ambient temperatures. Environmental Research, 1(1), 1-20. Ghavampour, E. Del Aguila, M. Vale, B. (2016). GIS mapping and analysis of behavior in small public spaces. Royal Geographic Society, 49(3), 349-358. Doi: 10.1111/area.12323 Giuffrida, L., Lokys, H., & Klemm, O. (2018). Assessing the effect of weather on human outdoor perception using twitter. International Journal of Biometeorology, , 1-12. doi:http://dx.doi.org.proxy2.cl.msu.edu/10.1007/s00484-018-1574-7 Gumprecht, B. (2007). The campus as a public space in an American college town. University of Massachusetts Press, 40-70. DOI: 10.1016/j.jhg.2005.12.001 Hajrasouliha, A. (2016). Campus score: Measuring university campus qualities. Landscape and Urban Planning, 158, 166-176. Doi:10.1016/j.landurbplan.2016.10.007 Kozlowski, Marek and Yusnani Mohd Yusof. 2016. "The Role of Urban Planning and Design in Responding to Climate Change: The Brisbane Experience." International Journal of Climate Change Strategies and Management 8 (1): 80-95. Lin, T. P. (2009). Thermal perception, adaptation and attendance in a public square in hot and humid regions. Building and Environment, 44(10), 2017-2026. doi:http://dx.doi.org.proxy2.cl.msu.edu/10.1016/j.buildenv.2009.02.004 Marcus, C. C., & Francis, C. (1990). People places: Design guidelines for urban open space. New York, N.Y: Van Nostrand Reinhold.Montgomery J. (1998). Making a city: Urbanity, vitality and urban design. Journal of Urban Design, 3, 93-116. Montigny, L. d., Ling, R., & Zacharias, J. (2012). The effects of weather on walking rates in nine cities. Environment and Behavior, 44(6), 821-840. doi:http://dx.doi.org.proxy2.cl.msu.edu/10.1177/0013916511409033 Siu, K.W. (2015). Flexible design principles: Street furniture design for transforming environments, diverse user’s, changing needs and dynamic interactions. Facilities, 33(9- 10), 588. Thorsson, S., Honjo, T., Lindberg, F., Eliasson, I., & Lim, E. (2007). Thermal comfort and outdoor activity in japanese urban public places. Environment and Behavior, 39(5), 660- 684. doi:http://dx.doi.org.proxy2.cl.msu.edu/10.1177/0013916506294937 Orum, A. (2009). Public man and public space in Shanghai today. City and community, 8(4), 369-389. 68 Peters, K. (2010). Social interactions in urban parks: Stimulating social cohesion?. Urban Forestry and Urban Greening, 9(2), 93-100. Doi: 10.1016/j.ufug.2009.11.003 Şatir, S., & Korkmaz, E. (2005). Urban open spaces with examples & the classification of urban furniture. A Z: ITU Journal of Faculty of Architecture, 2(1-2), 130-141. Siu Yu Lau, S., Gou, Z., Liu, Y. (2014). Healthy campus by open space design: Approaches and guidelines. Frontiers of architectural research, 3(4), 452-467. Doi: 10.1016/j.foar.2014.06.006 Verme, P. (2009). Happiness, freedom and control. Journal of economic behavior and organization, Pg. 146-161. Doi:10.1016/j.jebo.2009.04.008 Wang, B. (2001). Study of the Individualization of the Urban Plaza Space Environment. Hauzhong University of Science and Technology. Whyte, W.H. (19.80). The social life of small urban spaces. Washington, DC: the conservation foundation. Ye, P., Xv, X., Lin, S. (2013). Correlation study of urban plaza environment and behavior based on GPS. Applied Mechanics and Materials, 409-410. Doi: 10.4028/www.scientific.net/AMM.409-410.1026. Zacharius, J. Stathopoulos, T. Wu,H. (2004). Spatial behavior in San Francisco’s plazas, the effect of microclimate, other people, and environmental design. Environment and Behavior, 36(5), 638-658. Doi: 10.1177/0013916503262545. Zang, Hui. (2010). A Research on the City Leisure Square Space Dynamic Elements and Design Strategy. Dalian University of Technology. Doi: 10431880. 69