Street configuration and neighbourhood accessibility

Different street and block configurations are associated with different eras of development. (See Fig. 31.) As planning techniques changed, so too did the way street systems in new suburbs were laid out. With the arrival of the automobile in the 1920s and 1930s, the Victorian gridiron was deemed to be unsafe and was replaced by hierarchical networks of discontinuous streets intended to shield residential neighbourhoods from through traffic. Now these networks are generally believed to discourage walking and cycling and to undermine the cost-effective provision of public transit, thereby encouraging reliance on the automobile.

Fig. 31:Street patterns in different eras

Literature review

Southworth and Owens (1993) divide street systems into five patterns by era of development, ranging from the "gridiron" of prewar neighbourhoods to the curvilinear systems and cul-de-sac patterns found in postwar developments. They found that more recently developed subdivisions feature lower overall street length, fewer blocks, intersections, and access points to neighbourhood units, and more loops and cul-de-sacs. They suggest that lower street connectivity discourages walking and cycling.

Others have considered the relationship between street network configuration and travel behaviour. For 12 Seattle-area sites, Moudon et al. (1997) studied pedestrian network connectivity, route directness, and completeness of pedestrian facilities, which refers to the extent and distribution of pathways protected from vehicular traffic. The study found that sites classified as "urban" had much higher pedestrian traffic flows than suburban sites. The study concluded that while density, land use mix, and income are not sufficient to predict pedestrian travel volume, pedestrian network connectivity, route directness, and completeness of pedestrian facilities have a significant effect. A follow-up study by Hess et al. (1999) found that urban sites with small blocks and extensive sidewalk systems had, on average, three times the pedestrian volume of suburban sites.

These findings are reinforced by Riekko's (2005) aggregate study of the GTA, in which he found that transit use was positively correlated with smaller block sizes and gridded streets, although the relationship with population and dwelling unit density was stronger. All things being equal, the propensity to use transit was 150% higher in areas with gridded rather than curvilinear street networks, while a 10% increase in gross dwelling unit density increases the likelihood of transit use by 2%.

There is a lively debate on how best to operationalize neighbourhood accessibility in research. As noted, Southworth and Owens (1995) consider total road length, as well as the number of blocks, intersections, access points, loops, and cul-de-sacs for areas of equivalent size. In his review of the literature, Krizek (2003) proposes counting intersections and measuring average block size as indicators of accessibility. Criterion Planners' INDEX model (2004) proposes an "internal" street connectivity ratio, in which the total number of intersections is divided by the number of intersections plus cul-de-sacs. A higher number is understood to correspond to greater intra-neighbourhood connectivity. The model also includes an "external" street connectivity indicator -- the average distance between points of entry and exit to a neighbourhood. More data-intensive approaches quantify road length with and without sidewalks (Moudon et al. 1997; Hess et al. 1999). In a five-case study of how neighbourhood development patterns have changed over time, Knaap et al. (2005) found that in neighbourhoods built since 1940, internal connectivity progressively declined up to 1970 and then began to increase again. External connectivity, however, showed no temporal pattern across the five cases. Similar measures were employed by Weston (2002).

The salience of internal and external connectivity, and the distinction between them, is increased by the widespread adoption in planning practice of Perry's (1929) neighbourhood unit concept. To increase the safety of residents and especially children, he proposed replacing the extension of the gridiron with a hierarchy of streets that would define neighbourhoods. Through traffic would travel on arterial roads around each neighbourhood's edges. Within neighbourhoods, automobile traffic would be oriented locally, with narrow and curved streets serving to slow it down. The neighbourhood unit was to contain a population sufficient to support an elementary school -- approximately 5,000 to 6,000 people on 65 hectares. Postwar planning on such neighbourhood unit lines has been criticized for promoting automobile use at the expense of walking (Banerjee & Baer 1984).

Research questions

1. How do street network configurations differ by era of initial development?

2. Do the prewar study areas score higher in indicators of neighbourhood accessibility?

Findings

Changing street network configurations

Fig. 32 shows the street networks of all 16 study areas at the same scale. The street networks of the districts planned and built out the earliest -- Riverdale, Leaside, the two Oshawa cases, and Whitby -- feature gridded street patterns interrupted only by rivers, railroads, and large-scale land uses such as parks, schools, industrial uses, and shopping centres (which were inserted into the grid more recently). Inspired by garden suburb principles, Todd's 1912 plan for Leaside curves the grid, but the common elements -- narrowly spaced through streets and small blocks -- remain.

The 1960s-70s study areas display the influence of the neighbourhood unit concept and the use of looped streets and cul-de-sacs to slow down automobiles and discourage through traffic in residential areas. In each, there is a hierarchy of streets. The largest are arterials for through traffic, often corresponding to the original 2km (11/4 mile) surveyor's grid. Within these arterials, the area is divided into several neighbourhood units, each surrounding a school and a park. Collector streets loop through the area, connecting the neighbourhood units to each other and to the arterials. Within each neighbourhood unit, local streets, loops, and cul-de-sacs branch off the collector streets. In Milton, Bronte, Malvern, and Meadowvale, collectors subdivide the large areas defined by the arterials, cutting across from edge to edge. In Mississauga Valleys and the Peanut, the collector takes the form of a ring road.

The 1980s-90s study areas are transitional. Despite very different housing type mixes, Glen Abbey, Markham Northeast, and Vaughan feature full or partial ring-road systems similar to Mississauga Valleys. The neighbourhood units, however, are predominantly defined by linear natural heritage systems rather than parks and schools. Markham Northeast (1980s) and Cachet (early 1990s) bear strong similarities to the conventional suburbs of the 1960s and 1970s, although Cachet has a tighter network of collector roads. Richmond Hill and Vaughan, both developed in the 1990s, however, display a different approach. The central portions of Richmond Hill and Vaughan west of Jane St. feature grid systems and smaller blocks. These grid systems are internal to the neighbourhood units, however; they do not connect to the arterial roads.

Fig. 32: Study area street networks

Fig. 33: External connectivity (average distance in metres between entry/exit points)

Lower values indicate greater neighbourhood accessibility.

External street connectivity

Fig. 33 shows that the earlier a study area was developed, the more points of connection its internal street network has with surrounding arterial roads and, by extension, the shorter the average distance between points of entry and exit. While in the pre-1960 study areas the average distance between entry points is 282m, it is 773m and 662m in the 1960s-70s and 1980s-90s study areas, respectively.

Averages for the 1960s-70s and 1980s-90s groups are somewhat inflated by the existence of impassible features such as ravines, highways, and railway lines. For example, Meadowvale is bounded by a rail line (parallel to Highway 401) to the north and an industrial park to the east, while Bronte is bordered by a highway to the north and protected greenspace to the east. The remaining 1960s-70s study areas have values comparable to the 1980s-90s cases, suggesting that neotraditional urban design principles introduced since the 1980s have not increased external connectivity. This is as expected, given the disconnection of grids internal to neighbourhood units from arterials, and is consistent with the findings of Knaap et al. (2005).

Fig. 34: Road density (total road length in metres per developable hectare)

Higher values indicate greater neighbourhood accessibility.

Internal street connectivity

Three measures of internal connectivity were tested:

  • road density, expressed as linear road length per developable hectare;
  • intersection density, the number of intersections per developable square kilometre;
  • intersection frequency, the number of intersections per kilometre of road length.

The latter two exclude cul-de-sacs, which do not contribute to neighbourhood accessibility.

At over 125 metres per developable hectare, the road density values for pre-1960 study areas are among the highest in the sample. This is consistent with the fact that the pre-1960 cases, with their more tightly spaced street grids, have more road area. (See Fig. 23.) In general, the 1960s-70s neighbourhoods have lower road densities than the pre-1960 and post-1980 study areas. (See Fig. 34.)

Fig. 35: Intersection density (intersections per developable square kilometre)

Higher values indicate greater neighbourhood accessibility.

A similar pattern holds for intersection density. The pre-1960 study areas have more intersections per developable square kilometre (intersection density) than those developed later. (See Fig. 35.) Again, this is consistent with expectations. Riverdale and Old Oshawa have the highest intersection densities due to their grid systems and small block sizes. Bronte and Milton have low intersection densities due to the presence of superblocks of employment land. Cachet's value is low because of the large-lot "estate" subdivision in its northeast quadrant, which is served by looped streets. If these three low-value cases are set aside, the values for the remaining eight post-1960 study areas vary little, ranging from 42 to 48 intersections per square kilometre of developable land area.

On average, the pre-1960 group has a higher intersection frequency than the post-1980 group. The 1960s-70s group lies in between, with individual values comparable to those in the other groups. (See Fig. 36.) This is not entirely consistent with expectations -- one would have expected that the pre-1960 gridiron would score the highest and the 1960s-70s superblock developments the lowest. While the three oldest cases are among the highest, the 1960s-70s group contains both the highest and the lowest values, because of idiosyncrasies in the design of the individual study areas. Malvern's value is boosted by its many loops, which produce T-junctions. Bronte's value is low due to its large tract of employment land. If the two extreme cases are set aside, the values for the remaining post-1960 study areas vary little, ranging from 3.2 to 3.8 intersections per road kilometre.

Fig. 36: Intersection frequency (intersections per road kilometre)

Higher values indicate greater neighbourhood accessibility.

This convergence of most post-1960 study areas on a narrow band of values for intersection density and frequency indicates that the introduction of grid elements in some recently developed areas (Richmond Hill and Vaughan) has not been enough to produce scores equal to those of prewar areas.

One aspect of internal connectivity that this study could not address is the existence of or potential for non-street pathways and physical barriers to pedestrian travel. Many of the post-1960 study areas contain shopping malls and other employment areas. It is not known how accessible such areas are to nearby residential areas by foot or bicycle.

Composite indicator of neighbourhood accessibility

To gain an overall sense of the study areas' accessibility, their scores in each of the four indicators were ranked, and the resulting values summed. The results are shown in Fig. 37. In high-performing cases -- Old Oshawa, Riverdale, Whitby, and Leaside -- and low-performing cases -- Milton, Cachet, and Bronte -- the indicators tend to covary. In the middle band of composite scores -- Glen Abbey, Malvern, Mississauga Valleys, Richmond Hill, the Peanut, Vaughan, Oshawa West, Markham Northeast, and Milton -- covariation of the indicators is weak or absent. For example, while Malvern and Vaughan score poorly on external connectivity and road density, they have high intersection density and frequency scores. For Mississauga Valleys, the reverse is true. These findings are explained by the idiosyncrasies in neighbourhood design described previously.

Fig. 37: Ranking of study areas by neighbourhood accessibility indicator scores

Rankings by neighbourhood

accessibility indicator scores

External connectivitya

Road densityb

Intersection densityb

Intersection frequencyb

Composit score (sum)

Rank order by

composite score

Old Oshawa

2

1

1

3

7

1

Riverdale

1

4

2

2

9

2

Whitby

4

2

3

8

17

3

Leaside

3

7

4

4

18

4

Glen Abbey

7

11

8

5

31

5

Malvern

12

15

5

1

33

6

Mississauga Valleys

6

5

11

12

34

7

Richmond Hill

9

3

7

15

34

8

The Peanut

8

8

9

10

35

9

Vaughan

14

10

6

6

36

10

Oshawa West

5

9

13

11

38

11

Markham NE

11

6

12

13

42

12

Meadowvale

16

12

10

7

45

13

Milton

13

14

14

9

50

14

Cachet

10

13

15

14

52

15

Bronte

15

16

16

16

63

16

The lower the study area's score, the more accessible it is.
a. External connectivity scores are ranked from lowest to highest.
b. Internal connectivity scores are ranked from highest to lowest.

When ordered by era of initial development, the rankings present a clear pattern. With the exception of Oshawa West, whose results are skewed because of several large, campus-format land uses, the pre-1960 (in essence, prewar) cases are the most accessible. By contrast, the study areas developed later display no coherent pattern. The breakpoint in the scores is between pre- and postwar development patterns.

Summary of findings

1. How do street network configurations differ by era of initial development?

Consistent with general assumptions about how the planning of street networks has changed over time, study areas planned or built before the Second World War feature uniform grids and little differentiation between "major" and "minor" streets. The districts built after 1960, by contrast, were designed with street hierarchies intended to regulate traffic flow between and within relatively isolated neighbourhood units. The two late-1990s districts reintroduce grid elements within neighbourhood units, but their internal street systems largely do not connect to arterial roads.

2. Do the prewar study areas score higher in measures of neighbourhood accessibility?

As expected, the pre-1960 districts have higher external street connectivity than all subsequently developed areas. This may illustrate the enduring strength of the neighbourhood unit concept in planning practice, in which opportunities for traffic to infiltrate neighbourhoods bounded by arterial roads are strictly limited. This is true even in cases where neighbourhood units are laid out in grid systems because these grids are not connected to arterial roads.

Also as expected, the pre-1960 districts exhibit the highest degree of internal street connectivity. On average, the post-1980 districts score slightly higher than the 1960s-70s districts, perhaps indicating the abandonment of tower-in-the-park planning models and the influence of neotraditional planning ideas. These changes are not enough to produce scores rivalling those of the prewar grid-based study areas.

Implications for policy

Postwar neighbourhoods score lower on the accessibility indicators than prewar neighbourhoods. If more accessible street configurations do reduce automobile use in favour of walking and cycling -- a hypothesis that will be explored further in Section 2.7 -- this will require the connection of streets both within and between neighbourhood units. Greater Toronto's grid of arterial roads -- 2,000 metres on each side, enclosing 400 hectares -- is coarser than that in many other North American cities. In the Canadian prairies and American West and Mid-West, surveyors divided the land into smaller square-mile "sections" of 260 hectares. When the land is urbanized, the roadways bounding these survey units tend to become arterial through-streets. Although it may be an accident of history, Greater Toronto's coarser grain of arterials may frustrate connectivity and the potential for travel by means other than the automobile. Policies encouraging the subdivision of concession squares into a finer grid may promote more walking and cycling.