Urban area types and spatial distribution of pedestrians: Lessons from Tel Aviv

Computers, Environment and Urban Systems 55 (2016) 11–23

Contents lists available at ScienceDirect

Computers, Environment and Urban Systems journal homepage: www.elsevier.com/locate/ceus

Urban area types and spatial distribution of pedestrians: Lessons from Tel Aviv Yoav Lerman ⁎, Itzhak Omer Department of Geography and Human Environment, Tel-Aviv University, Tel-Aviv, Israel

a r t i c l e

i n f o

Article history: Received 17 February 2015 Received in revised form 12 September 2015 Accepted 23 September 2015 Available online xxxx Keywords: Pedestrian movement Space syntax Spatial analysis Modernism

a b s t r a c t This study examines the role of two urban area types – traditional and contemporary – with regard to pedestrian movement volume and distribution. This study focuses on four dimensions of urban areas which have potential influence on pedestrian movement: (i) a spatial dimension based on road network structure; (ii) a functional dimension of land uses such as retail fronts; (iii) a physical dimension of road sections; and (iv) a demographic dimension of population and employment densities. Four research areas in Tel Aviv are examined and each of these areas is divided to two adjacent sub-areas — a traditional sub-area and a contemporary one. The aim is to clarify: (i) the character of urban areas that were created following different urban design paradigms; (ii) the relative contribution of the spatial, functional, physical and demographic dimensions to pedestrian movement in urban areas of different types. The findings show significant differences between adjacent traditional and contemporary sub-areas. Specifically, traditional sub-areas have higher levels of spatial connectivity and retail fronts distribution as well as higher pedestrian movement volume. The spatial dimension has the strongest overall connection to pedestrian movement, and particularly for traditional sub-areas, while the physical dimension has the strongest connection to pedestrian movement for the contemporary sub-areas. © 2015 Elsevier Ltd. All rights reserved.

1. Introduction Recently, there has been a growing awareness of the importance of pedestrian movement in the urban realm, while recognizing its environmental, health and safety advantages (Frank et al., 2006; Jacobsen, 2003). This change is occurring after a period of modernist planning which put a big emphasis on making private vehicle movement relatively fast and frictionless (Murrain, 2002). As the connection between urban and transportation planning is becoming clearer (Forsyth & Southworth, 2008) the effects on pedestrian movement call for deeper investigation. A significant body of literature (Hillier, Penn, Hanson, Grajewski, & Xu, 1993; Raford & Ragland, 2006; Read, 1999) has been accumulated with regard to pedestrian movement in traditional environments, though relatively few studies addressed pedestrian movement in contemporary environments. The general objective of this study is to examine the relationship between two different urban area categories – traditional versus contemporary – and pedestrian movement volume in the city of Tel Aviv. Generally speaking, road network structure can be divided to traditional on the one hand and contemporary on the other according to geometric and configurational attributes such as the number of intersections, number and size of blocks, connectivity and centrality (Forsyth & Southworth, 2008; Marshall & Garrick, 2010; Murrain, 2002; Peponis, Allen, Haynie, ⁎ Corresponding author. E-mail address: [email protected] (Y. Lerman).

http://dx.doi.org/10.1016/j.compenvurbsys.2015.09.010 0198-9715/© 2015 Elsevier Ltd. All rights reserved.

Scoppa, & Zhang, 2007). Essentially, areas planned under contemporary planning doctrines combine the functional hierarchy of streets with the neighborhood unit concept, which results in dendritic and disconnected urban networks (Marshall, 2005). In contrast, traditional urban areas enhance accessibility to residential and non-residential land uses through their early-grid network. Furthermore, a study conducted in the US found that areas built after 1950 have a higher average length of street block, a lower number of streets and accordingly are more car oriented with low walkability level compared with areas that were built earlier (Peponis et al., 2007). Some of the neighborhoods in Tel Aviv are new neighborhoods that were established following a modern planning approach, while other neighborhoods in Tel Aviv are old neighborhoods that were established following traditional urban planning, Hence, due to the varying types of urban neighborhoods in Tel Aviv, we assume that the case of the Tel Aviv is ideal for investigating the implications of different planning approaches on pedestrian movement. Moreover, this study compares adjacent areas, traditional and contemporary neighborhoods, which are similar in socio-economic parameters (see below), yet differ in their urban type. This study is based on empirical research in four selected areas in the city of Tel Aviv, and aims to clarify: (i) the character of the built environment of urban areas that were created following different urban design paradigms; and (ii) the impact of the built environment characteristics on pedestrian movement distribution in the city's street network. Specifically, pedestrian movement volume in this study reflects an hourly

12

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

average on a typical weekday in the afternoon. Such an investigation may contribute to improving the evaluation of pedestrian movement distribution in urban areas that are already built and those that will be planned and built in the future, as well as to understanding how to create built environments which encourage pedestrian movement. Furthermore, combining physical elements with spatial and functional elements is not common in the related literature and this study aims to extend the understanding on the interplay between these built environment dimensions. 1.1. Street network's spatial configuration and pedestrian movement Many studies dealing with pedestrian movement volume in urban space rely on street network analysis by employing the configurational approach, primarily within the conceptual framework of space syntax (Hillier et al., 1993; Jiang, 2009a; Raford & Ragland, 2006). This theoretical framework is based on a topological-visual analysis of the street network called the axial map, defined as the smallest set of straight axial lines covering the urban street network. The analysis of the topological relationship between the axial lines produces a number of indices of the street grid (Hillier, 1996a) that describe the centrality of individual axial lines such as: connectivity, integration, and choice. Connectivity denotes the number of directly linked axial lines for a given axial line. Integration indicates the closeness of an axial line to all other axial lines in the system by computing the shortest topological distance of the respective line from every axial line in the road network, while the choice measure reflects the likelihood of through-movement. The integration and choice indices can be calculated on a global level (taking into account the entire axial map), or on a local level according to different distance radii. A higher radius indicates a larger extension of the network. For example, a local integration index with a radius of 3 (r = 3) reflects a line topological proximity to its nearby axial lines which includes in the calculation only those lines, which are up to two topological steps away from it. This local index is common in studies that employ space syntax techniques to evaluate pedestrian movement volume (Jiang, 2009a; Raford, 2003; Read, 1999). For a detailed description of the integration and choice indices, which correspond to the graph theory-based measures closeness and betweenness, respectively (see: Hillier & Iida, 2005, pp. 481–483). In addition to analysis based on axial lines, it is possible to do network analysis on a finer scale through the use of segments. By employing this type of resolution, angular distance (least cumulative angle measure) and metric distance can be considered besides topological distance. Studies comparing the relevance of these three distance measures to urban movement show that angular distance have the highest correlations compared to the other two (Hillier & Iida, 2005; Turner, 2007). While the space syntax approach relies on a secondary representation of the road network, other analyses are based on the primal representation of the road network map (i.e. using the GIS road center lines map). A study conducted in Hong Kong compared topological analysis based on street names to analysis using space syntax measures. In that study, street name based variables achieved higher statistical correlations to vehicular movement compared to axial based variables (Jiang & Liu, 2009). Thus, street based variables should also be considered in pedestrian movement analyses. 1.2. Other factors influencing pedestrian movement In addition to the influence of the road network on pedestrian movement it was found that the land use layout has an impact on pedestrian distribution in urban space. Several studies found that retail has a strong influence on pedestrians' presence in its vicinity (Liu & Griswold, 2009; Zook, Lu, Glanz, & Zimring, 2012). Other studies combine land use and configurational factors to further understand the relationship between the built environment and pedestrian movement and their respective contribution to an area walkability level (Lerman, Rofè, & Omer,

2014; Ozer & Kubat, 2007; Raford & Ragland, 2006). These studies show that spatial structure tends to have greater connection than retail to pedestrian movement whereas combining them together give better correlation values to pedestrian movement volume. For example, in such combined models for pedestrian movement prediction in Boston, the most significant variable was ‘space syntax integration’, while other variables that improved the model were proximity to public transit stations and proximity to tourist attractions (Raford & Ragland, 2006). The physical structure pertaining to the road sections may also have an impact on pedestrian volume. Desyllas and Duxbury (2000) found that sidewalk width correlates with pedestrian movement volume so that excessive congestion of pedestrians on a sidewalk leads to lower pedestrian volume due to slower movement. Another study (Desyllas, Duxbury, Ward, & Smith, 2003) concluded yet again that there is a significant connection between pedestrian volume on a given sidewalk and that sidewalk width. Lastly, demographic characteristics may also be associated with pedestrian movement volume. Several studies demonstrated correlations between pedestrian movement and population and employment densities (Liu & Griswold, 2009; Kim, Ko, & Lee, 2013), while another study that combined spatial analysis with demographic data showed the demographic attributes had also significant influence accounting for 45% out of the model regression correlation (Raford, 2003). Based on the scientific knowledge about the distribution of pedestrian movement, the current study focuses on the role of urban character in the relationship between built environment features and pedestrian movement volume in urban space. Four main dimensions of the urban environment which have potential influence on pedestrian movement are considered: (i) a spatial dimension which is based on the road network structure; (ii) a functional dimension of land uses such as retail fronts and public transit stops; (iii) a physical dimension of road sections; and (iv) a demographic dimension of residential and employment densities. The complete list of variables used in this study is specified in Section 2. 1.3. Research questions and hypotheses The research questions are: 1. How are the differences between traditional and contemporary urban areas reflected in terms of spatial, functional, physical and demographic dimensions? 2. How does the different urban character correlate with pedestrian volume movement and distribution at the level of city streets? 3. What is the relative contribution of the spatial, functional, physical and demographic urban dimensions to pedestrian movement? The spatial dimension refers to the spatial relation between urban objects i.e., streets, open spaces and buildings. The functional dimension refers to the functional content of the buildings. The physical dimension refers to the physical dimensions of the urban objects i.e., road width and sidewalk width. The demographic dimension refers to population and employment densities. These are the relevant urban dimensions for pedestrian movement, as discussed in the pedestrian movement research. 4. What are the relationships among these urban dimensions in urban areas of different types? Following previous studies (Marshall & Garrick, 2010; Peponis et al., 2007) we hypothesize that the differences between traditional and contemporary urban areas would be consistent and substantial, and would lead to a different pattern of walking in them. Based on previous research we also hypothesize that the spatial dimension would have the most significant affect on pedestrian movement compared to the other urban dimensions examined in this research. Furthermore, according to the space syntax theory (Hillier, 1996b, 1999) we assume that the spatial dimension affects land use distribution, especially retail, and would therefore correlate with the functional dimension. Studies on the physical road structure are rarely connected to spatial analysis (Desyllas & Duxbury,

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

2000; Desyllas et al., 2003). Therefore, we do not make assumptions on the possible relationships between the physical dimension and the other urban dimensions or on its connection to pedestrian movement. We also do not refer in this study to environmental factors such as noise, climate comfort, and shade which may impact pedestrian movement distribution (e.g. Almeida, 2007). 2. Method 2.1. Research area Four research areas were examined in the city of Tel Aviv. Each of these areas is composed of two adjacent sub-areas that differ in their urban character — one sub-area is traditional and the other is contemporary. Fig. 1 shows the location of the different areas in Tel Aviv, while Fig. 2 gives a closer aerial look on each of the research areas. Table 1 gives the

13

size and name of each of the areas and sub-areas. The research areas were selected from the more central parts and less central parts of the city. There are a few common features that distinguish the traditional subareas from the contemporary sub-areas. First, the traditional sub-areas have all been mostly built prior to 1936 (and planned or unplanned prior to that), while the contemporary sub-areas have been planned and built afterwards in the 1940s and 1950s. Second, the traditional sub-areas tend to have a denser and more connected road network compared to the contemporary sub-areas (see the aerial photos in Fig. 2). Tel Aviv has undergone a major shift in its planning regime between these two periods as explained in Marom (2014, p. 1915): “Throughout the 1930s–1940s, Mayor Israel Rokach and Chief Municipal Engineer Yacov Shifman Ben-Sirah – both educated as civil engineers and espousing an ‘engineerial’ habitus – envisioned and enacted a project of urban modernization and territorial expansion

Fig. 1. The four research areas locations in the city of Tel Aviv. Each area is marked with a number: (1) Ibn Gabirol; (2) Florentine; (3) Yad Eliyahu–Hatikvah; and (4) Shapira–Kiriyat Shalom.

14

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

Fig. 2. A closer look on the four research areas and the division between the traditional and contemporary sub-areas. (1) Ibn Gabirol; (2) Florentine; (3) Yad Eliyahu–Hatikvah; and (4) Shapira–Kiriyat Shalom.

through dozens of municipal plans. Their approach thus differed from the laissez-faire mode of development under Mayor Dizengoff in the 1920s, which envisioned a city of bourgeois entrepreneurship and high culture.” Following is a description of each of the research areas and sub-areas characteristics. It should also be noted that while differing in urban character, the adjacent sub-areas are similar in terms of socio-economic status

(Omer & Goldblatt, 2012, Fig. 1a, p. 180). Area number 1, (named Ibn Gabirol after the street that divides its two sub-areas) consists of western and eastern parts. The western sub-area was built during the 1930s along a master plan made in the 1920s by Sir Patrick Geddes. The eastern area was built along more modernistic lines with more open spaces, less mix of land uses, wider major roads and less density during the 1950s as part of the “East Tel-Aviv” Plan. Area number 2 is called Florentine after

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

15

Table 1 The 15 research samples and the number of road sections surveyed in each of them. Area name

Sub-area

Urban type

All areas combined Traditional sub-areas combined Contemporary sub-areas combined Ibn Gabirol West East

Traditional Contemporary

Residential Industrial

Traditional Contemporary

Hatikvah Yad Eliyahu

Traditional Contemporary

Shapira Kiriyat Shalom

Traditional Contemporary

Florentine

Yad Eliyahu–Hatikvah

Shapira–Kiriyat Shalom

the neighborhood it encompasses. The distinction between its sub-areas is based on significant street network and land use differences. The north-eastern sub-area was built during the 1920s and 1930s and combines rather high residential density with an eclectic commercial ground floor. The south-western part consists mainly of heavy industry and has relatively low density of housing and was mostly built during the 1940s. Area number 3 is made of Hatikvah neighborhood (south of the divider) and Yad Eliyahu neighborhood. Hatikvah neighborhood was built during the 1930s in an informal way without a master plan. Yad Eliyahu neighborhood was established in 1945 following a modernistic approach with low level of mix of uses and no commercial streets. Area number 4 consists of Shapira (north of the divider) and Kiriyat Shalom neighborhoods. While Shapira neighborhood started developing in the 1920s on small lots and in a somewhat independent fashion, Kiriyat Shalom was established much later in 1951 following a modernistic curvilinear pattern.

Calculation of the space syntax attributes was based on the axial map of the entire city of Tel Aviv. The calculation was performed using the DepthMap software.1 The following variables were computed: Axial-based variables (topological space syntax): (1) Connectivity. (2) Integration — global and local (with r = 3). (3) Choice — global and local (with r = 3).

1 Depthmap is a spatial network analysis software available at: https://github.com/ SpaceGroupUCL.

215 102 99 51 24 24 52 25 23 60 28 28 52 25 24

556.57 244.89 311.68 161.95 63.87 98.08 67.45 35.31 32.14 175.96 82.15 93.81 151.20 63.55 87.65

2.2.2. Functional variables Since our spatial analysis is based on high-resolution analysis at the level of street segment only functional properties which can be applied to a street segment and were found most relevant for pedestrian movement were computed. The following three variables were computed: (1) Commercial fronts – this variable was calculated based on an actual field survey of all street segments in the research areas. Each street segment was given a value of 0, 1 or 2 depending on the amount of commercial fronts in it (retail on two sides, one side or none).

In this section the study variables are specified according to the four urban dimensions examined — spatial, functional, physical and demographic.

(1) Street connectivity — the number of streets a given street intersects with. (2) Street length. (3) Ratio of street length to its connectivity.

Size (hectares)

Segment-based variables (geometric and metric space syntax): For the segment analysis we have computed both geometric and metric mean depth and choice values. All segment-based variables were computed with different metric radii as follows: 250 m, 500 m, 750 m, 1000 m, 1250 m, 1500 m, 1750 m, 2000 m, 2500 m, 3000 m, 4000 m, 5000 m and radius n (entire system).

2.2. Built environment features

2.2.1. Spatial variables Spatial variables include street based variables and space syntax variables. The spatial extent of the map used for the spatial analysis was set at Tel Aviv city limits. The following three variables are based directly on the street network map excluding road sections that have no name:

No. of sections

Two variables were calculated to represent proximity to public transit.2 The analysis was conducted using a buffer of 100 m from each side of the axial lines – which generally represent the center line of the streets – to capture all points in a given street. The buffer of 100 m was determined after several examinations of various distances. A smaller buffer did not capture all the points located along the street, while a larger buffer included in many cases points that are located along neighboring streets. (2) Number of bus stations within a 100 m radius. (3) Number of bus lines within a 100 m radius.

2.2.3. Physical variables Four variables have been computed for each of the road segments in the research areas. These variables are based on a field survey conducted by the authors across all the road segments in the research areas. Specifically, the variables of ‘carriage way level’ and ‘sidewalk width’ were determined for each road segment as follows:

2

Based on official Israel governmental map website: http://govmap.gov.il/.

16

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

(1) Carriage way level — quantifies the difficulty for crossing the road on foot: 1 – pedestrian only path; 2 — minor road; 3 — medium road with light bus traffic; 4 — major road that is somewhat risky to cross; and 5 — a busy road that can only be crossed at specified intersections with traffic lights. (2) Sidewalk width — quantifies the width of the sidewalk according to the number of people that can walk abreast on it: 0 — no sidewalk; 1 — one person can walk abreast; 2 — two persons can walk abreast; and 3 — three or more persons can walk abreast. The value itself for a given road section is the sum of the sidewalk widths of both sides, so a road with two wide sidewalks has a value of 6. (3) Road width — the sum of carriage way level and sidewalk width. (4) Ratio of sidewalk width to carriage way level.

2.2.4. Demographic variables Three demographic variables were computed for each research area and sub-area based on the 2008 Israel census3 for residential densities and the NTA database4 for employment densities. These variables were not computed at the granularity of street sections. The following variables were computed: (1) Gross residential density — the ratio between the number of residents to area size (in hectares). (2) Gross employment density — the ratio between the number of employment places to area size (in hectares). (3) Activity density — the sum of residential and employment densities. (4) Ratio of residential to employment densities.

2.3. Pedestrian movement volume counts Our dependent variable is the average number of pedestrians per hour in a given street segment. Measurement points in the research areas were selected to represent a range of different centrality measures and distribution of land uses. The method of gate counts was used in each survey point (i.e. the number of pedestrians moving through a specific point in a street segment). The counts were done for 5 min every hour for 5 h in each point on major roads, and for 10 min every hour on minor roads (to reduce the variance due to lower movement volume). In each area the survey was done simultaneously for all survey points. The surveys took place on sunny weekdays between the hours 3 pm and 8 pm. During these hours commerce still operates while foot traffic is boosted at the end of the workday and school. The survey took place in months of optimal weather when school season is active. The entire research sample included 215 road segments where pedestrian movement volume was sampled in all four research areas combined. This sample was divided into 15 different sub-samples according to the different geographical areas that the study dealt with (Table 1). Different aggregation levels were used to infer differences among various urban scales and contexts. The samples include eight sub-areas, four research areas, two urban design paradigms (traditional subareas combined and contemporary sub-areas combined) and the entire sample. These 15 research samples were used as the basis for statistical analyses that were conducted to find the most significant variables associated with pedestrian movement.

3 The Israeli Central Bureau of Statistics 2008 census website: http://cbs.gov.il/census/ census/main_mifkad08_e.html. 4 NTA is Metropolitan Mass Transit System Ltd., which builds Tel Aviv Metro light-rail system.

3. Results 3.1. Differences between traditional and contemporary built environments Analysis of the research areas shows that urban areas of different types differ in substantial and consistent ways in their built environment features (see Table 2). The traditional sub-areas tend to have a more connected road network, and consist of a higher presence of commercial land use, compared to their contemporary counterparts. In particular, these differences stand out when examining sub-areas that follow different design paradigms and are located next to each other, as examined in this study. To test for significance of the differences in the spatial structure between the traditional and contemporary sub-areas, independent t-tests were done for the groups of streets and axial lines in each of the four research areas as well as for the combined samples of all contemporary and traditional sub-areas (Table 3). The t-tests of the combined sample show that there are statistically significant differences for all the spatial measures between traditional and contemporary areas. The t-tests for the specific research areas show that in some areas these spatial differences pertain to the local scale and in others they are related to the global scale of the entire city. As for differences in functional properties, it can be clearly seen that retail fronts are more prevalent in the traditional sub-areas (Fig. 3 in Section 3.2). The traditional sub-areas contain three to eight times more retail fronts in terms of percentage out of the entire available street fronts. Commercial streets (streets where there are at least a few consecutive segments with retail on both sides) are completely absent from all the contemporary sub-areas, while in each of the traditional sub-areas there is at least one commercial street. Examination of the demographic data shows that the traditional areas have a higher activity density (residential and employment densities combined) than their contemporary counterparts. In the first and second research areas the difference in the activity density is higher than 50% while in the third and fourth research areas these differences are 10–20% only. One more finding regarding the ratio between residences and employment places suggests that the traditional areas tend to have a more balanced mix of uses, e.g., in the fourth research area there are eight residents per one employment place in the contemporary sub-area versus 3.5 residents per one employment place in the traditional sub-area. 3.2. Pedestrian movement distribution Significant disparities in pedestrian movement volume distribution were found among the research areas and sub-areas. Table 4 presents the average pedestrian movement volume per road section in the research areas and sub-areas. Higher volume of pedestrian movement was observed in the traditional sub-areas compared to their contemporary counterparts. Moreover, even in the traditional sub-area where the lowest pedestrian movement volume was observed, the average movement volume per road segment was higher still than that in the contemporary sub-area with the highest pedestrian movement volume (even though the activity density in that traditional sub-area is lower than the activity density of that contemporary sub-area). Fig. 3 shows the distribution of pedestrian movement volume (using average per hour values) in the survey points in all research areas. Pedestrian movement volume was found to have a heavy-tailed distribution, meaning that in most of the survey points a low volume was measured, while in a few survey points a high volume of pedestrian movement was measured. This finding is compatible with previous studies (Jiang, 2009b) that show that a minority of roads carry high movement volume, and in most of the road the movement volume is low. To visually emphasize the survey points where high volume of pedestrian movement was recorded an algorithm that breaks a heavytailed distribution to head and tail categories was used (Jiang, 2013).

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

17

Table 2 Built environment measures for the four research areas and eight sub-areas (Trad. = traditional; Cont. = contemporary). Ibn Gabirol Whole area

Trad.

Florentine

Out of the 26 points that were categorized as high pedestrian movement volume points, 15 are inside traditional sub-areas, 6 are on roads that serve as borders between sub-areas and only 5 are inside contemporary sub-areas. 3.3. Correlations between built environment features and pedestrian movement volume distribution This section presents the squared correlation values (R2) of pedestrian movement volume with regard to various research variables. Since pedestrian movement has a heavy-tailed distribution it was normalized using the log function (with the natural logarithm). All choice and connectivity variables also have heavy-tailed distributions and were also normalized using the log function. Table 5 presents the research variables from all the different variable categories that have the highest average correlation in relation to the 15 research samples or the highest correlation coefficient for one or more research samples. All correlations are significant with P b 0.01 unless mentioned otherwise. Overall, we used six categories to distinguish among the variables as the spatial dimension was divided to four subcategories: spatial- street name; spatial-axial; spatial-geometric; and spatial-metric. The last two variable categories pertain to the functional and physical variables.

Cont.

Shapira–Kiriyat Shalom

Street connectivity–average 6.12 6.73 5.61 7.26 7.94 7.37 5.08 5.20 5.13 5.06 6.09 4.15 Axial connectivity–average 5.63 6.51 5.51 7.89 8.38 8.07 6.32 6.79 6.11 4.79 5.57 4.51 Global integration–average 1.16 1.17 1.15 1.07 1.10 1.04 1.03 1.00 1.06 0.93 1.03 0.87 Local integration (r = 3)–average 2.48 2.69 2.36 2.75 2.85 2.71 2.60 2.63 2.62 2.11 2.33 1.99 Percent of commercial fronts (out of all fronts) 0.11 0.14 0.05 0.37 0.60 0.07 0.05 0.08 0.01 0.03 0.06 0.01 Percent of road length with dual commercial fronts 0.09 0.12 0.03 0.35 0.59 0.05 0.05 0.07 0.00 0.02 0.05 0.01 Percent of pedestrian paths (length) 0.09 0.05 0.12 0.04 0.04 0.04 0.24 0.17 0.32 0.18 0.10 0.25 Percent of major roads (length) 0.07 0.00 0.13 0.03 0.00 0.08 0.04 0.04 0.01 0.05 0.08 0.03 Residential density (per hectare) 173.51 233.28 134.59 122.36 198.24 39.00 130.71 133.90 127.92 100.87 101.52 100.40 Employment density (per hectare) 90.98 109.82 78.70 230.67 263.80 194.27 27.85 33.32 23.07 19.22 28.45 12.53 Activity density (residential + employment) 264.49 343.10 213.29 353.04 462.04 233.28 158.56 167.22 150.99 120.09 129.96 112.93 Residential to employment ratio 1.91 2.12 1.71 0.53 0.75 0.20 4.69 4.02 5.54 5.25 3.57 8.01

2. Medium pedestrian movement volume (less than 408 but more than 191.7 pedestrians per hour on average) — 47 points accounting for 22% of the survey. 3. Low pedestrian movement volume (less than 191.7 pedestrians per hour on average) — 142 points accounting for 66% of the survey.

Trad.

Whole area

Spatial–street Spatial–axial Spatial–axial Spatial–axial Functional Functional Physical Physical Demographic Demographic Demographic Demographic

1. High pedestrian movement volume (over 408 pedestrians per hour on average) — 26 points accounting for 12% of the survey.

Whole area

Yad Eliyahu–Hatikvah

Measure description

This method breaks a heavy-tailed distribution in a deterministic fashion and captures the underlying hierarchy of the data. This is done by partitioning all the data values around the mean into two parts and continuing the process iteratively for the values (above the mean) in the head until the head part values are no longer heavy-tailed distributed. Applying this scheme on the 215 survey points resulted with the following three categories (used in Fig. 3):

Cont.

Whole area

Measure type

Trad.

Cont.

Trad.

Cont.

The spatial variable street connectivity which is based on the street network has the highest average squared correlation value for the 15 research samples (average R2 of 0.46) and also has the highest correlation values for five different research samples (the most of all the research variables). This spatial variable tends to have stronger correlations with pedestrian movement for the traditional sub-areas. As for the space syntax related variables — the geometric variable of mean depth with a radius of 1250 m has the highest average correlation value among them. The geometric variables tend to have higher squared correlation values than metric variables. It is also worth noting that no axial variable achieves the highest correlation value for even a single sample. The functional variable with the highest squared correlation value to pedestrian movement is commercial fronts (average R2 of 0.34), which achieves the highest correlation coefficient for a single research sample. This functional variable also tends to have stronger correlations for the traditional sub-areas samples. As for the physical variables — the highest squared correlation values are achieved with the variable of road width (average R2 of 0.38), which achieves the highest correlation coefficients for three research samples. This physical variable has stronger correlations to pedestrian movement for the contemporary sub-areas samples rather than for the traditional sub-areas. Another physical variable with a strong connection to pedestrian movement is sidewalk width (average R2 of 0.32) which was found to have positive correlation to pedestrian movement volume in a previous research (Desyllas et al., 2003). Examination of this variable correlation to pedestrian movement reveals differences among the research areas. For example, for the first research area (Ibn Gabirol) there is a markedly higher correlation (0.55) than for the second research area (Florentine with 0.27). This phenomenon may point to difference in the planning actions taken in these areas so that the sidewalks in the first area where better fitted for the pedestrian movement volume compared to the sidewalks in the second area. Essentially, pedestrian movement volume may have an impact on the road sections themselves and the width of the sidewalks and carriage ways. The next step was to conduct a multiple regression analysis for each of the 15 research samples. The dependent variable was log value of the average pedestrian volume movement per hour, while the independent variables included the entire list of variables described previously. The multiple regressions were done using a stepwise method. The regression correlation coefficients (R2) ranged between 0.54 and 0.94

Table 3 Statistical significance of independent t-tests for several variables of sub-areas. Variable name

All zones together

Ibn Gabirol

Florentine

Yad Eliyahu–Hatikvah

Shapira–Kiriyat Shalom

Street connectivity Axial connectivity Global integration Local integration (r = 3)

P b 0.05 P b 0.01 P b 0.01 P b 0.01

Not significant Not significant Not significant P b 0.01

Not significant Not significant P b 0.01 Not significant

Not significant P b 0.05 P b 0.01 Not significant

Not significant Not significant P b 0.01 P b 0.01

18

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

Fig. 3. Observed pedestrian movement volume at the surveyed points in the research areas. Also displayed is the distribution of commercial fronts. (1) Ibn Gabirol; (2) Florentine; (3) Yad Eliyahu–Hatikvah; and (4) Shapira–Kiriyat Shalom.

Table 4 Average pedestrian movement volume per road section in each of the research areas and sub-areas. Research area name

Whole area avg.

Traditional sub-area avg.

Contemporary sub-area avg.

Ratio

Ibn Gabirol Florentine Yad Eliyahu–Hatikvah Shapira–Kiriyat Shalom

250.5 180.3 188.2 149.6

280.7 255 243 179.2

174.6 82.9 108.4 126.2

1.6 3.1 2.2 1.4

0.07†† 0.32

0.12†† 0.31 0.10†† 0.34 0.27 0.33 0.50 0.32 0.26 0.34 0.32 0.38

0.38 0.23 0.29

0.45 0.31 0.33

0.18 0.30 0.41

0.40 0.55 0.60

0.43 0.43 0.59

0.25† 0.62 0.67

0.38 0.27 0.30

0.50 0.14†† 0.42 0.13† 0.21† 0.08† 0.24† 0.34 0.13

0.00†† 0.05†† 0.22† 0.15

0.00††

0.00††

0.05†

0.11†

0.03†† 0.01†† 0.52

0.48

0.10†† 0.11

0.16 0.32 0.33 0.17† 0.22

0.15

0.26

0.05†

0.30

0.32

0.13†

0.45

0.00†† 0.30

0.12† 0.24 0.33 0.30

0.27

0.31

0.17

0.46

0.44

0.40

0.43

0.50

0.10†† 0.35

0.21† 0.59 0.47

0.14†† 0.50 0.29 0.38 0.37 0.37

0.37

0.43

0.21

0.52

0.37 0.34 0.59

0.41

0.32

0.26

0.41 0.31 0.41

0.13†† 0.60 0.32 0.17 0.29

0.28

0.36

0.15

0.36

0.35 0.18† 0.43

0.35

0.21†

0.14†† 0.30

0.00††

0.05†† 0.23† 0.41 0.26 0.16 0.22 0.36 0.46 0.50 0.35 0.12†† 0.16

Functional Physical Physical

Spatial–Metric

Spatial–Metric

Spatial–Geometric

Spatial–Geometric

Spatial–Geometric

Local integration (r = 3) Geometric choice (750 m radius) Geometric mean depth (750 m radius) Geometric mean depth (1250 m radius) Global geometric mean depth Metric choice (2500 m radius) Metric mean depth (5000 m radius) Commercial fronts Sidewalk width Road width

0.31 0.24

0.26 0.15

0.30 0.20

0.15 0.08

0.56 0.24

0.48 0.34

0.28 0.31 0.24 0.23† 0.00†† 0.49

0.20† 0.50

0.54

0.27† 0.59 0.46 0.43 Street connectivity

Spatial–street name Spatial–axial Spatial–Geometric

0.46

0.43

0.53

0.37

0.59

0.78

0.38

0.45

0.19†

0.35 0.64 0.45

0.22† 0.16††

Cont. Trad. Whole area

Shapira–Kiriyat Shalom

Cont. Trad.

Yad Eliyahu–Hatikvah

Whole area Cont. Trad. Whole area Cont. Trad.

Florentine Ibn Gabirol

Cont. Whole sub-areas area Trad. sub-areas Entire survey All samples avg. value Variable name Variable type

Table 5 Squared correlation coefficients (R2) for selected research variables for the 15 research samples (Trad. = traditional; Cont. = contemporary). Shown here are variables that have the higher correlation coefficient to one or more of research samples and variables that have the highest average correlation value in their category. All the correlations are significant with P b 0.01 unless otherwise mentioned. The symbol (†) signifies that the correlation is significant with P b 0.05 and the symbol (††) signifies that the correlation is not significant. The highest correlation coefficient for each survey sample is written in bold.

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

19

Table 6 (a) Prevalence of built environment variable categories in the 15 regression models. (b) Prevalence of specific research variables in the 15 regression models. (a) Variable type

No. of regressions

No. as lead variable

Spatial–all Spatial–street Spatial–axial Spatial–geometric Spatial–metric Functional Physical

14 8 3 7 10 11 9

9 4 0 4 1 1 5

(b) Variable name

Variable type

No. of regressions

As lead variable

Commercial Fronts Street connectivity Metric mean depth (5000 m radius) Road width Sidewalk width Ratio of street length to connectivity Global integration Geometric mean depth (750 m radius) Geometric mean depth (no radius) Geometric choice (no radius) Metric mean depth (1000 m radius) Metric mean depth (3000 m radius) No. of bus stations within 100 m radius Street length Axial connectivity Geometric mean depth (250 m radius) Geometric mean depth (1750 m radius) Geometric choice (500 m radius) Geometric choice (750 m radius) Geometric choice (3000 m radius) Metric mean depth (1250 m radius) Metric mean depth (1750 m radius) Metric mean depth (4000 m radius) Metric mean depth (no radius) Metric Choice (250 m radius) Metric Choice (3000 m radius) No. of bus lines within 100 m radius Ratio of sidewalk to carriage way Carriage way level

Functional Street Metric

10 7 5

1 4 1

Physical Physical Spatial–street

4 4 2

2 3 0

Spatial–axial Spatial–geometric

2 2

0 1

Spatial–geometric

2

1

Spatial–geometric

2

0

Spatial–metric

2

0

Spatial–metric

2

0

Functional

2

0

Spatial–street Spatial–axial Spatial–geometric

1 1 1

0 0 0

Spatial–geometric

1

0

Spatial–geometric

1

1

Spatial–geometric

1

1

Spatial–geometric

1

0

Spatial–metric

1

0

Spatial–metric

1

0

Spatial–metric

1

0

Spatial–metric

1

0

Spatial–metric

1

0

Spatial–metric

1

0

Functional

1

0

Physical

1

0

Physical

1

0

compared to a range of 0.35–0.78 when using single variables in bivariate correlations. In addition, the regressions correlation coefficients tend to be higher for the traditional sub-areas samples compared to the contemporary sub-areas samples. Out of all the research variables 29 appear in the 15 different regression runs (all regressions are statistically significant with P b 0.01). Table 6a shows the prevalence of the different types of built environment variables in the regressions, while Table 6b shows the specific

20 Table 7 Rankings for each of the 29 variables that take place in the multiple regressions (value of 1 means the variable is the leading variable in the regression) for the 15 survey samples. The specific contribution of each variable to the regression correlation coefficient (R2) is given in parenthesis. All regressions are statistically significant with P b 0.01. Ibn Gabirol

Yad Eliyahu–Hatikvah

Shapira–Kiriyat Shalom

Trad. sub-areas

Cont. sub-areas

Whole area

Trad.

Cont.

Whole area

Trad.

Cont.

Whole area

Trad.

Cont.

Whole area

Trad.

Cont.

Street 0.67 6

Street 0.76 6

Phys. 0.54 4

Phys. 0.84 4

Street 0.88 3

Phys. 0.84 3

Metric 0.74 4

Func. 0.94 7

Geo. 0.62 2

Geo. 0.82 5

Geo. 0.83 4

Street 0.84 4

Physical 0.65 4

Geo. 0.86 4

Phys. 0.67 2

2 (0.15) 1 (0.43)

2 (0.13) 1 (0.53)

3 (0.05)

4 (0.02)

2 (0.07) 1 (0.78) 3 (0.03)

3 (0.06)

2 (0.19) 3 (0.07)

2 (0.2)

2 (0.16) 1 (0.6)

5 (0.02)

3 (0.03)

3 (0.04) 4 (0.03) 5 (0.01)

2 (0.11) 1 (0.67)

4 (0.03) 1 (0.52) 2 (0.12)

1 (0.5) 5 (0.06)

2 (0.17) 1 (0.64) 2 (0.11)

1 (0.3)

1 (0.34)

1 (0.59)

5 (0.02) 3 (0.07) 1 (0.5)

1 (0.6)

3 (0.08)

2 (0.16)

3 (0.04) 2 (0.1)

4 (0.03)

4 (0.04) 4 (0.02)

4 (0.06) 3 (0.06) 4 (0.06) 4 (0.05) 4 (0.07) 1 (0.5) 1 (0.49) 3 (0.11) 3 (0.11) 3 (0.07) 4 (0.05)

6 (0.01) 6 (0.02) 6 (0.03) 2 (0.12) 2 (0.17) 7 (0.02)

2 (0.08)

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

Lead variable type Regression Coefficient (R2) No. of variables in regression Variable name Commercial Fronts Street connectivity Metric mean depth (5000 m radius) Road width Sidewalk width Ratio of street length to connectivity Global integration Geometric mean depth (750 m radius) Geometric mean depth (no radius) Geometric choice (no radius) Metric mean depth (1000 m radius) Metric mean depth (3000 m radius) No. of bus stations within 100 m radius Street length Axial connectivity Geometric mean depth (250 m radius) Geometric mean depth (1750 m radius) Geometric choice (500 m radius) Geometric choice (750 m radius) Geometric choice (3000 m radius) Metric mean depth (1250 m radius) Metric mean depth (1750 m radius) Metric mean depth (4000 m radius) Metric mean depth (no radius) Metric Choice (250 m radius) Metric Choice (3000 m radius) No. of bus lines within 100 m radius Ratio of sidewalk to carriage way Carriage way level

Florentine

Entire survey

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

21

Table 8 Correlation coefficients (R2) among the leading built environment variables and pedestrian volume movement for the 15 research samples (Trad. = traditional; Cont. = contemporary). The symbol (†) signifies that the correlation is significant with P b 0.05 and the symbol (††) signifies that the correlation is not significant. (a) Leading spatial variable, street connectivity. (b) Leading functional variable, commercial fronts. (c) Leading physical variable, road width. Ibn Gabirol Variable name (a) Pedestrian movement volume Commercial fronts Road width (b) Pedestrian movement volume Street connectivity Road width (c) Pedestrian movement volume Street Connectivity Commercial fronts

Florentine

Yad Eliyahu–Hatikvah

Shapira–Kiriyat Shalom

All samples avg. value

Entire survey

Trad. sub-areas

Cont. Whole sub-areas area

Trad.

Cont.

Whole area

Trad.

Cont.

Whole area

Trad.

Cont.

Whole area

Trad.

Cont.

0.46

0.43

0.53

0.37

0.59

0.78

0.43

0.38

0.45

0.19†

0.46

0.45

0.64

0.35

0.59

0.27†

0.17 0.37

0.17 0.34

0.25 0.25

0.05† 0.42

0.21 0.61

0.22† 0.02†† 0.23 0.57 0.66 0.19

0.41 0.17† 0.06†† 0.23†

0.10† 0.37

0.12†† 0.07†† 0.18 0.30 0.54 0.36

0.36 0.29

0.06†† 0.36

0.34

0.38

0.45

0.18

0.40

0.43

0.50

0.50

0.41

0.21†

0.17 0.14

0.17 0.09

0.25 0.14

0.05† 0.11

0.21 0.23

0.22† 0.02†† 0.23 0.36 0.10†† 0.06††

0.41 0.17† 0.08†† 0.23†

0.10† 0.09†

0.12†† 0.07†† 0.18 0.20† 0.08†† 0.11†

0.36 0.06†† 0.17† 0.07††

0.38

0.29

0.33

0.41

0.60

0.59

0.67

0.24†

0.13

0.26

0.27

0.41

0.37 0.14

0.34 0.09

0.25 0.14

0.42 0.11

0.61 0.23

0.57 0.36

0.66 0.19 0.10†† 0.06††

0.37 0.09†

0.30 0.20†

0.54 0.36 0.08†† 0.11†

variables prevalence in the regressions. The spatial variables take part in 14 out of 15 regression, the functional variables appear in 11 and the physical variables in 9. In two regression formulas variables of the spatial dimension only are used, while for the rest of samples a combination of variables from different dimensions resulted in higher correlation coefficient. The most prevalent variable is the functional variable of commercial fronts, which takes part in 10 different regression models. The spatial variable of street connectivity appears in the highest number of regression models as the leading variable (4 regressions) and has the second highest prevalence (7 regressions). Table 7 presents the specific contributions of all research variables in the multiple regressions. 3.4. Correlations between leading variables from different dimensions Following the results presented above, the relationships among variables from different dimensions were explored, to understand the way the built environment features impact one another. We chose one representative variable from each dimension, the variable that had the highest average squared correlation coefficient to pedestrian movement. The three variables thus examined are: the spatial variable of street connectivity, the functional variable of retail fronts and the physical variable of road width. This examination relied on bivariate correlations among these three variables across the 15 research samples. Table 8(a–c) shows these correlation coefficients including the correlation coefficient for pedestrian movement volume. The squared correlation value between the leading spatial variable and the leading functional variable is 0.17 on average, and shows significant disparities between traditional and contemporary sub-areas. For the four traditional sub-areas the average squared correlation value between these two variables is 0.28, while for the four contemporary subareas this figure is 0.08. The connection between the leading spatial variable and the leading physical variable is stronger and has an average squared correlation coefficient of 0.37. The physical variable has a stronger connection to the spatial variable in the contemporary sub-areas where the average correlation is 0.45 compared to average correlation of 0.30 in the traditional sub-areas. The spatial–physical squared correlation values tend to be much higher than the spatial–functional values for the contemporary sub-areas (0.45 compared to 0.08), while these average values are similar for the traditional sub-areas (0.30 compared to 0.28). This finding

0.25†

0.38

0.30

0.14†† 0.42

0.34

0.06†† 0.23† 0.08†† 0.23†

0.12†† 0.31

0.33

0.47

0.29 0.36 0.17† 0.07††

indicates that the spatial configuration of traditional sub-areas relates to both physical and functional features, while in the contemporary sub-areas there is a much stronger congruence with the physical structure than with functional features. This is a result of the stricter modernistic planning approach to transportation and land use. The connection between the leading functional variable and the leading physical variables is weaker than either of these variable connections to the spatial variable. The average squared correlation value for the functional–physical connection is 0.14. This connection tends to be stronger in the traditional sub-areas where the average is 0.20 compared to an average of 0.12 in the contemporary sub-areas. The squared correlation values between pedestrian movement and these three variables are higher on average than those among these research variables themselves. While for the spatial variable and the physical variable these differences are rather minor, for the functional variable of retail fronts we find that it had a much higher correlation to pedestrian volume movement than to the other research variables. The average correlation between the functional variable and pedestrian movement is 0.34 (compared to 0.17 with the spatial variable which has the highest correlation from among the research variables) and it is higher than the correlation to the other research variables in 14 out of the 15 research samples. This finding resonates with Hillier's notion of the economic multiplication (Hillier, 1996b) generated by a reciprocal connection between pedestrian movement and retail distribution. 4. Discussion The research findings indicate that the urban environment features are associated with pedestrian distribution at the street level. However, this influence may be differential in accordance with the urban area type. The spatial dimension is the dimension with the highest level of association with pedestrian movement and in particular makes the highest contribution in the multiple regressions employed in this research. This finding is consistent with results of previous studies (Jiang, 2009a; Raford & Ragland, 2006). However, the research variable that has the most significant statistical connection to pedestrian movement is the street based connectivity, which is not based on space syntax analysis. This finding may point to that fact that pedestrian movement in urban space is not based wholly on visual perception as is often claimed in space syntax related research (Hillier & Iida, 2005)

22

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23

Hence, pedestrian movement distribution impacts the functional dimension and in particular retail fronts distribution. The results also show that pedestrian movement distribution may affect the physical dimension through alterations to the road sections so as to make them more convenient for pedestrians where their volume is high. Fig. 4 presents a diagram of the connections among the urban area type, built environment dimensions and pedestrian movement. This diagram is an extension of the diagram presented by Hillier et al. (1993, p.31, Fig. 3) for describing the connections between spatial and functional dimensions (named ‘A’ for Attraction in Hillier's parlance) and pedestrian movement. Fig. 4 shows the dominance of the urban area type over the built environment connections. Furthermore, the reciprocal connections between pedestrian movement and the functional and physical dimensions are shown, while the spatial dimension has a one-way impact on the other dimension and on pedestrian movement. Essentially, pedestrian movement may serve as an intermediary between urban dimensions so that physical dimension impacts on pedestrian movement may lead to changes in the functional dimension (and vice-versa). 5. Conclusions

Fig. 4. The diagram of connections between urban area type, built environment dimensions and pedestrian movement. C is spatial configuration, A is functional dimension (attraction in Hillier et al., 1993), P is the physical structure and M is pedestrian movement. The urban area type is at the root of the diagram. The spatial configuration affects the functional dimension, the physical structure and pedestrian movement but is not effected by any of them. The functional and physical dimensions have reciprocal connections with pedestrian movement.

but may also be related to cognitive space represented by street names, as found concerning vehicle movement (Jiang & Liu, 2009; Turner, 2007). Regarding the functional dimension of the built environment, it is shown that commercial land uses are related to greater pedestrian movement in their vicinity as found in previous studies (Liu & Griswold, 2009; Zook et al., 2012). However, a finding that is not common in the scientific literature is the significant relationship found between pedestrian movement distribution and the physical dimension. The physical dimension has stronger association with pedestrian movement in the contemporary sub-areas, where the spatial and functional dimensions tend to have weaker connections to pedestrian movement relative to the traditional sub-areas. This connection is not necessarily the one that was expected by the modernistic planning doctrine under which pedestrians were supposed to congregate in the inner neighborhood spaces and leave the major roads for motorized transportation (Hebbert, 2005; Marshall, 2005, pp. 45–70). However, higher pedestrian volume movement was observed along major roads of contemporary sub-areas, where vehicular movement travels at higher speed. As to the demographic dimension, it was found that the traditional sub-areas tend to have higher densities of activities as well as a more balanced mix between residences and employment places compared to the contemporary sub-areas. However, the differences in densities are not correlated with the differences in pedestrian volume movement, i.e. traditional sub-areas of relatively low density had higher pedestrian movement volume than contemporary sub-areas that are denser. Analysis of the correlations between built environment's features and pedestrian movement distribution revealed a strong relationship and co-structuration between the two. The spatial structure serves as the basis for a certain pedestrian movement and along the roads where there are high levels of pedestrian movement commercial activity may form. In turn, the commercial activity helps to intensify pedestrian movement in its vicinity in what is described as the economic multiplication effect of the city movement economy (Hillier, 1996b).

The spatial dimension of the urban environment has a prominent connection to pedestrian movement compared to the other dimensions. The combination of the spatial configuration and the pedestrian movement that formulates on it may lead to changes in the functional and physical dimensions. These findings serve to substantiate the claim that the spatial structure has a significant effect on various urban dynamics (Hillier, 1996a; Hillier et al., 1993). Therefore, when dealing with urban planning it is advisable to dedicate much thought to the planning of the road network, which is inherently static and difficult to change. This observation is compatible with other diagnoses in this field (Hillier, 1996b; Lerman et al., 2014; Marshall, 2005). The limited number of research areas suggests the results presented may be too particular. Further studies involving the examination of pedestrian movement in contemporary and traditional areas and the examination of the differences in the urban features between these two urban design paradigms may help in validating these results. Moreover, the most important research variable was found to be street connectivity and this finding suggests further research should be conducted using street based spatial variables. However, street based variables may have limitations caused by arbitrary circumstances. To overcome this, natural streets based on good continuity principle may also be used in further research (Jiang & Liu, 2009). Finally, further research should examine the connection between physical road sections and pedestrian volume movement. Modernistic urban planning assumed that a clear road hierarchy would reduce the number of pedestrians using the major roads and would keep them walking on internal neighborhood roads (Hebbert, 2005) and this may not be the case after all. References Almeida, M.D. (2007). The importance of shade as a strategy to foster walking in summer. Proceedings of the eighth International Walk 21 Conference, Toronto, 1–4 October, 2007. Desyllas, J., & Duxbury, E. (2000). Planning for movement: Measuring and modelling pedestrian flows in cities. RICS Conference, London (Retrieved from: http://discovery.ucl. ac.uk/233/). Desyllas, J., Duxbury, E., Ward, J., & Smith, A. (2003). Demand modeling of large cities: An applied example from London. UCL Centre for Advanced Spatial Analysis. Forsyth, A., & Southworth, M. (2008). Cities afoot — Pedestrians, walkability and urban design. Journal of Urban Design, 13(1), 1–3. http://dx.doi.org/10.1080/ 13574800701816896. Frank, L.D., James, S.F., Terry, C.L., Chapman, J.E., Saelens, B.E., & Bachman, W. (2006). Many pathways from land use to health: Associations between neighborhood walkability and active transportation, body mass index, and air quality. Journal of the American Planning Association, 72(1), 75–87. http://dx.doi. org/10.1080/01944360608976725. Hebbert, M. (2005). Engineering, urbanism and the struggle for street design. Journal of Urban Design, 10(1), 39–59. http://dx.doi.org/10.1080/13574800500062361.

Y. Lerman, I. Omer / Computers, Environment and Urban Systems 55 (2016) 11–23 Hillier, B. (1996a). Space is the machine. Cambridge, MA: Cambridge University Press. Hillier, B. (1996b). Cities as movement economies. Urban Design International, 1(1), 41–60. http://dx.doi.org/10.1057/udi.1996.5. Hillier, B. (1999). Centrality as a process: accounting for attraction inequalities in deformed grids. Urban Design International, 4(3), 107–127. http://dx.doi.org/10.1038/ udi.1999.19. Hillier, B., & Iida, S. (2005). Network effects and psychological effects: A theory of urban movement. In A. van Nes (Ed.), Proceedings of the Fifth International Space Syntax Symposium (pp. 475–490). Delft: Techne Press. Hillier, B., Penn, A., Hanson, J., Grajewski, T., & Xu, J. (1993). Natural movement: Or, configuration and attraction in urban pedestrian movement. Environment and Planning B: Planning and Design, 20http://dx.doi.org/10.1068/b200029. Jacobsen, P.L. (2003). Safety in numbers: More walkers and bicyclists, safer walking and bicycling. Injury Prevention, 9(3), 205–209. http://dx.doi.org/10.1136/ip.9.3.205. Jiang, B. (2009a). Ranking space for predicting human movement in an urban environment. International Journal of Geographical Information Science, 23(7), 823–837. http://dx.doi.org/10.1080/13658810802022822. Jiang, B. (2009b). Street hierarchies: A minority of streets account for a majority of traffic flow. International Journal of Geographical Information Science, 23(8), 1033–1048. http://dx.doi.org/10.1080/13658810802004648. Jiang, B. (2013). Head/tail breaks: A new classification scheme for data with a heavytailed distribution. The Professional Geographer, 65(3), 482–494. http://dx.doi.org/ 10.1080/00330124.2012.700499. Jiang, B., & Liu, C. (2009). Street-based topological representations and analysis for predicting traffic flow in GIS. International Journal of Geographical Information Science, 23(9), 1119–1137. http://dx.doi.org/10.1080/13658810701690448. Kim, D., Ko, J., & Lee, Y. (2013). Estimating pedestrian traffic volume: A preliminary analysis. Proceedings of the Eastern Asia Society for Transportation Studies, 9. (pp. 154). Taipei: Department of Transportation (09–12 September 2013). Lerman, Y., Rofè, Y., & Omer, I. (2014). Using space syntax to model pedestrian movement in urban transportation planning. Geographical Analysis, 46(4), 392–410. http://dx. doi.org/10.1111/gean.12063. Liu, X., & Griswold, J. (2009). Pedestrian volume modeling: A case study of San Francisco. Yearbook of the Association of Pacific Coast Geographers, 71(1), 164–181. http://dx.doi. org/10.1353/pcg.0.0030.

23

Marom, N. (2014). Planning as a principle of vision and division: A bourdesian view of Tel Aviv's urban development, 1920s–1950s. Environment and Planning A, 46(8), 1908–1926. http://dx.doi.org/10.1068/a46301. Marshall, S. (2005). Streets and patterns. London: Spon Press. Marshall, W.E., & Garrick, N.W. (2010). Street network types and road safety: A study of 24 California cities. Urban Design International, 15, 133–147. http://dx.doi.org/10. 1057/udi.2009.31. Murrain, P. (2002). Understand urbanism and get off its back. Urban Design International, 7, 131–142. http://dx.doi.org/10.1057/palgrave.udi.9000081. Omer, I., & Goldblatt, R. (2012). Urban spatial configuration and socio-economic residential differentiation: the case of Tel Aviv. Computers, Environment and Urban Systems, 36(2), 177–185. http://dx.doi.org/10.1016/j.compenvurbsys.2011.09.003. Ozer, O., & Kubat, A.S. (2007). Walking initiatives: A quantitative movement analysis. In A.S. Kubat, Ö. Y. Ertekin, I. Güney, & E. Eyüboglu (Eds.), Proceedings of the Sixth International Space Syntax Symposium, 91. (pp. 1–13). Istanbul: ITU Faculty of Architecture. Peponis, J., Allen, D., Haynie, D., Scoppa, M., & Zhang, Z. (2007). Measuring the configuration of street networks. In A.S. Kubat, Ö. Y. Ertekin, I. Güney, & E. Eyüboglu (Eds.), Proceedings of the Sixth International Space Syntax Symposium, 002. (pp. 1–16). Istanbul: ITU Faculty of Architecture. Raford, N. (2003). Looking both ways: Space syntax for pedestrian exposure forecasting and collision risk analysis. In J. Hanson (Ed.), Proceedings of the Fourth International Space Syntax Symposium, 51. (pp. 1–16). London: University College London. Raford, N., & Ragland, D.R. (2006). Pedestrian volume modeling for traffic safety and exposure analysis: Case of Boston, Massachusetts. Transportation Research Board 85th Annual Meeting Compendium of Papers (Paper #06–1326 (CD-ROM)). Read, S. (1999). Space syntax and the Dutch city. Environment and Planning B: Planning and Design, 26, 251–264. http://dx.doi.org/10.1068/b260251. Turner, A. (2007). From axial to road-centre lines: A new representation for space syntax and a new model of route choice for transport network analysis. Environment and Planning B: Planning and Design, 34(3), 539–555. http://dx.doi.org/10.1068/b32067. Zook, J.B., Lu, Y., Glanz, K., & Zimring, C. (2012). Design and pedestrianism in a smart growth development. Environment and Behavior, 44, 216–234. http://dx.doi.org/10. 1177/0013916511402060.