Content uploaded by Anne Vernez Moudon
Author content
All content in this area was uploaded by Anne Vernez Moudon on Sep 16, 2015
Content may be subject to copyright.
94
■
Transportation Research Record 1818
Paper No. 02-3748
Planning tools were developed for local and state agencies to identify
locations with latent demand for pedestrian travel that are currently
underserved with pedestrian infrastructure. Prior research in the Puget
Sound showed that approximately 20% of the suburban population
lives in dense, compact areas with latent demand for pedestrian travel.
The tools are designed to enable agencies to target capital investments
in nonmotorized infrastructure to areas with the highest potential for
pedestrian trips. They are a first step toward delineating suburban
pedestrian zones. After a review of existing methodologies to identify
areas with pedestrian travel demand, two tools were developed that use
geographic information system software. One tool benefits from high-
resolution parcel-level data with specified land use attributes. The other
tool, however, relies on commonly available census block data and aer-
ial photography. It is more labor intensive than the first tool and re-
quires familiarity with reading urban form and development patterns.
The tools identify locations with potential for pedestrian travel based on
two attributes. First, the locations contain land uses that are function-
ally complementary, that is, commonly linked by travel. The land uses
are dense residential development (travel generators) and retail areas
and schools (travel attractors). Second, these land uses are also spatially
complementary, that is, sufficiently close to each other to be linked by
walking.
This paper reviews methods and presents tools developed to iden-
tify suburban areas that have potential for pedestrian travel. It is
based on previous research that showed that many suburban areas
designed for car travel are dense, mixed in use, and compact enough
to support pedestrian travel (1, 2). The tools help to identify these
types of areas to maximize the benefits of new capital investments
in pedestrian infrastructure.
BACKGROUND
Pedestrian travel is second only to automobile use as a mode of trip
making, yet its share of total travel is very small and continues to
decrease—from 10% to less than 6% of all trips in the United States
during the last 25 years (3). Other industrialized countries have a
far larger share of nonmotorized trips—11% in Canada, 20% in the
United Kingdom, and 49% in Sweden. Despite low trip rates, inter-
est in supporting pedestrian travel in U.S. metropolitan areas has
recently increased for several reasons. First, transportation policies
now seek to increase modes of travel other than the single-occupant
vehicle to alleviate traffic congestion and to improve environmen-
tal quality. Pedestrian travel serves these objectives because, given
proper conditions, it can substitute for short vehicle trips. Second,
the promotion of pedestrian travel may have a broad impact on
people’s choice of travel mode because walking trips are com-
monly used to access motorized modes, especially transit. Therefore,
poor walking environments may deter people from using transit.
Third, growing evidence shows that pedestrian-friendly places are
also desirable and safe environments that add to the quality of com-
munities. Finally, growing concern about high levels of physical
inactivity point to walking and biking as healthful activities with
popular appeal (4).
Given proper support, walking appears to remain a strong alter-
native to driving. Distances driven continue to be short—27% of all
trips are less than 1.6 km (1 mi), and an additional 13% are less than
3.2 km (2 mi) (5, 6). Public transit trips are increasing in several
metropolitan areas where supporting infrastructure investments
have been made (7).
At the same time, much work needs to be done to reverse the im-
pact of policies and practices that have supported automobile travel at
the expense of other modes. Fifty years of producing environments
that lack pedestrian infrastructure means that, today, more than 60%
of the U.S. population in large metro areas lives in environments that
discourage or even prevent people from walking. These environments
have been termed pedestrian-hostile (8). Policies have also led to lim-
ited budgets for infrastructure for nonmotorized travel at all levels of
government. For example, only 1% of the state of Washington’s trans-
portation safety budget is directed to pedestrian infrastructure
improvements—a disproportionately low figure considering that
pedestrian fatalities consist of 12% to 15% of all vehicle-related fatal-
ities. Similarly, the share of federal transportation funds targeted to
pedestrian travel remains minuscule (www.tea21.org/guide/charts/
2000.htm, www.walkinginfo.org/insight/fact_sheets/index.htm).
Yet policies have begun to shift. For example, federal policy state-
ments on transportation system design have directed state and local
agencies to incorporate pedestrian facilities into all projects unless
exceptional circumstances exist (9). Also, budgets for nonmotor-
ized infrastructure have consistently increased at many levels of
government over the last decade. Given the magnitude of the prob-
lem, however, these budgets remain insufficient to support signifi-
cant increases in nonmotorized travel. Effective resource allocation
at all levels of government is essential, requiring strategic planning
Pedestrian Location Identification Tools
Identifying Suburban Areas with Potentially High
Latent Demand for Pedestrian Travel
Anne Vernez Moudon, Paul Mitchell Hess,
Julie M. Matlick, and Nicholas Pergakes
A. Vernez Moudon, P. M. Hess, and N. Pergakes, Department of Urban Design
and Planning, University of Washington, Box 355740, Seattle, WA 98195.
J. M. Matlick, Community Partnerships Program, Washington State Department
of Transportation, 310 Maple Park Avenue SE, Olympia, WA 98504.
and sophisticated tools. Specifically, investments in nonmotorized
transport should be carefully targeted and prioritized to areas that
have the highest potential for substantial increases in nonmotorized
travel demand. The tools presented here are intended to aid in this
process.
CHARACTERISTICS OF DEMAND FOR
PEDESTRIAN TRAVEL
Data on pedestrian travel in the United States are limited, making it
difficult to target facility investments to where they may be most
needed (10). For example, Antonakos reports on data from the
Nationwide Personal Transportation Survey and finds that non-
motorists are likely to have the following characteristics: to be under
21 or over 65 years of age, unemployed, and without a driver’s
license and to live in a central city in a household with no auto-
mobile (11). National-level data therefore indicate that people who
walk do so only from necessity. In contrast, local-level data suggest
that walking is common for people of many age groups, with dif-
ferent types of personal characteristics, and in many types of urban
and suburban settings. On one hand, differences in mode-share exist
between metropolitan areas—the result of different periods of de-
velopment and related transportation infrastructure, transit service,
and local behaviors (12–14). On the other hand, marked differences
in behavior also exist within metropolitan areas. For example, walk-
ing trips account for about 20% of trip making by residents in some
medium-density, upper-income neighborhoods of Seattle—against
a low 2.5% as the average of King County, which includes Seattle
(15). In general, therefore, the few studies available that use dis-
aggregated data show that people other than the poor, the young,
and the old walk, and they do so in places other than the densest
urban centers.
Research for this study found substantial numbers of people walk-
ing in suburban areas that had not previously been identified as areas
with nonmotorized travel (1). The areas, termed suburban clusters,
do not correspond to recognized suburban downtowns or centers (16).
Instead, they are concentrations of medium-density multifamily devel-
opments in proximity to medium- and small-sized retail areas (2). Sub-
urban clusters have land use patterns that can support walking, but
they have poor pedestrian infrastructure. Infrastructure deficiencies,
including few sidewalks, difficult road crossings, and indirect
pedestrian routes, likely limit the number of people that walk in these
places. Neighborhoods with the same density and basic land use mix
but with extensive pedestrian infrastructure were shown to have three
times as many pedestrians (17, 18). These differences in walking rates
are an indicator of latent demand for pedestrian travel in the suburban
clusters.
Suburban clusters are common and suggest that a large latent
demand exists for pedestrian travel in suburban areas. Twenty per-
cent of the suburban population in the Puget Sound region (with a
total of 3.3 million around Seattle) lives in suburban clusters, an un-
expectedly high number considering recent growth and the preva-
lence of sprawling land use patterns. The result of common planning
and development practices, suburban clusters are also likely to be com-
mon in the postwar areas of other large metropolitan regions (2). By
applying Puget Sound figures to metropolitan regions with more than
1 million people, as many as 15 million to 20 million people already
may be living in suburban areas with land use patterns in which walk-
ing could be used as an alternative mode of travel for some trips.
Moudon et al. Paper No. 02-3748 95
However, these are places with little pedestrian infrastructure and
high latent demand for pedestrian trips.
These studies suggest that funding for nonmotorized travel must
look beyond the limited population of pedestrians living in estab-
lished older central cities and identify suburban populations living
in areas with potentially high rates of walking. The following two
sections review the state of the art in theory and methods to locate
and estimate demand for pedestrian travel.
DETERMINANTS OF WALKING
The choice to walk results from three determining factors (4):
•
Personal factors, such as income, age, health, ethnicity, educa-
tion, social and environmental attitudes toward the different means
of transportation, and type of employment;
•
Environmental factors, such as pedestrian-supportive land use
types, intensity, and mix; transportation facility characteristics; cli-
mate and weather; and topographical conditions; and
•
Trip characteristics, including trip length and purpose.
These determinants of travel interact in often complex ways to affect
mode choice. For example, a wealthy person may choose to drive a
short distance to work in downtown Manhattan but will walk 2 km
(1.2 mi) to a friend’s house in the country.
The steps required to identify the environmental conditions that
support walking and to locate latent demand for pedestrian travel are
reasonably well understood. First, walking as a mode of transport is
limited by the characteristics of the trip. Today, walking is practical
as a mode of transport for short and very short distances. This means
that the distance between trip origin and destination must be small.
The value given to walkable distances typically is derived from the
perceived cost of alternative modes of transport in functionality
(cost and time), safety, or comfort. A walking distance that is widely
applied is 0.4 to 0.8 km (0.25 to 0.5 mi), although it has little empir-
ical verification and has not been refined to include variations based
on location or personal determinants.
Second, pedestrian trips require complementary origin and desti-
nation. Land uses at trip origin and destination must functionally
complement each other as generators and attractors of travel. Appro-
priate generators and attractors include dense residential or employ-
ment areas as well as specific land uses such as shops, parks, and
schools. Many of the existing approaches used to measure the rela-
tionships between origins and destinations are imprecise or even
erroneous.
A third criterion for locating areas with potential for pedestrian
travel demand addresses the infrastructure or the facilities needed
for safe and comfortable walking. Continuous sidewalks and trails
and safe street crossings are readily used indicators of pedestrian-
supportive facilities. Although research has shown that the pres-
ence of appropriate facilities appears to be necessary to support
significant volumes of pedestrians (1), such facilities are never suf-
ficient to generate significant volumes of pedestrian trips. In other
words, the lack of a sidewalk system will deter many people from
walking, but the presence of a sidewalk system will not attract
pedestrians unless it connects appropriate origins and destinations.
This suggests that capital budgets for pedestrian infrastructure
should be targeted exclusively to places where such origins and
destinations are found. (The single exception to this rule is in cases
where new vehicle infrastructure is being developed; then non-
motorized facilities should always be included unless exceptional
circumstances exist.)
METHODOLOGICAL CONSIDERATIONS
A wide range of methods and tools has been developed to locate and
estimate pedestrian travel demand (4). Most of these methods need
to be empirically verified. One barrier to verification is the lack of
local data on actual pedestrian trip making mentioned earlier (10).
Another problem is that few estimation methods address the issue
of the latent demand for walking. Many estimate pedestrian volumes
based on projections from existing mode share and do not consider
the potential for latent demand. Others use an estimated percentage
of automobile trips that may be substituted, depending on the dis-
tance for biking or walking between origin and destination. Finally,
several of the methods are based on the concept of level of service
and relate demand estimation to infrastructure supply or pedestrian
delay. As shown earlier, however, the assumption that “they’ll come
if we build them” (4, p. 3) is valid only in those locations where
there are short distances between functionally complementary ori-
gins and destinations.
The pedestrian potential and deficiency indexes in use in the Port-
land, Oregon, metropolitan area are some of the most sophisticated
methods developed to date to identify and rank areas with potential
demand for pedestrian travel (18). The development of these
indexes follows earlier work done on the pedestrian environment fac-
tor (PEF), which was developed for use in modeling regional travel
(8). PEF was based on an adapted Delphi method of ranking fore-
cast analysis zones for sidewalk availability, ease of street crossing,
connectivity of street and sidewalk systems, and terrain. The pedes-
trian potential index (PPI), on the other hand, first attempts to locate
areas that have the appropriate land uses to support pedestrian travel.
A pedestrian deficiency index (PDI) then complements PPI. PDI
identifies and ranks deficiencies in the pedestrian infrastructure that
likely decrease the potential demand for pedestrian travel. PPI and
PDI therefore address both the potential of an area and its short-
comings in supporting pedestrian travel. Significantly, they also make
a distinction between land use conditions and infrastructure that do
or do not cater to pedestrian travel.
Portland’s PPI is similar in purpose to the tools described here. It
is useful to discuss PPI to introduce the approach taken to structure
tools in this study. PPI has two applications. One is relatively sim-
ple and is meant to be used in local planning applications. It contin-
ues to rely on the adapted Delphi method used in PEF. The other
requires land use data and comes closest to the methods and tools
presented in this paper.
The first simple application of PPI starts by identifying areas where
schools, parks, and neighborhood retail are present and assigns
points or weights to these land uses. This technique, also used in
level of service approaches (19–21), is referred to as that of prox-
imity factors. It assumes that these land uses attract pedestrian
travel. Although the identification of attractors is an important ele-
ment of any method, PPI does not identify the generators that may
or may not be located around the attractors. Yet this is important.
For example, an elementary school surrounded by 5,000 residents
living within 0.5 km likely will generate many more pedestrian trips
than will an elementary school surrounded by 1,000 residents within
0.5 km, all else being equal. The presented methods attempt to
address this issue.
96 Paper No. 02-3748 Transportation Research Record 1818
The second application of PPI relies on land use data to measure
land use intensity, mix, street connectivity, average parcel size, and
slope of terrain. This application of PPI measures environmental
factors more accurately than the first application, but it still has some
problems. First, PPI relies on spatial units of data that are too large
to capture actual development patterns, especially in suburban areas
(2, 22). Second, it is not clear that PPI adequately captures the land
use mix that supports pedestrian travel. It measures land use in both
residential and employment density without examining the particu-
lar mixes of activity that may support walking. Employment density
alone, for example, is not a good proxy for retail and service uses.
Although they may be attractors of pedestrian travel, these uses also
tend to have low employment densities. Further, the model assumes
that pedestrian travel will increase as both employment and resi-
dential densities increase. The reality of relationships between the
mix of uses is considerably more complex. High levels of pedestrian
activity do not necessarily occur in areas that have both medium or
high employment and housing. Usually, such activity takes place in
areas that have either medium to high employment densities or
medium to high residential densities. This is because in most U.S.
cities, residential and employment densities do not covary within
areas that are small enough for walking. For example, high employ-
ment densities are found in areas of office development, which typ-
ically do not have high- or even medium-density residential areas
within walkable distances. (There are exceptions, such as Manhat-
tan and the city centers of San Francisco, Boston, and, perhaps,
Philadelphia.) U.S. cities do have many areas of medium- to high-
density residential development, which are found in close proxim-
ity to retail, but these neighborhood retail uses have comparatively
low employment densities. Conversely, retail uses with high
employment densities, such as suburban shopping malls, are typi-
cally isolated (in pedestrian terms) from residential areas. Thus real-
world areas of mixed use typically include a dominant land use—
residential or employment—and other land uses. For areas to be
supportive of pedestrian travel these other land uses must be func-
tionally and spatially complementary with the dominant land use.
Whether residential or for employment, the dominant land use is the
prime generator of travel. Functionally, complementary land uses
are those that are likely to be linked by a trip from the dominant
land use. For example, schools and industrial uses are not comple-
mentary, as they are unlikely to generate many pedestrian trips
between them, but residential and retail, or employment and retail,
are complementary. Functionally complementary land uses are at-
tractors of pedestrian travel from the dominant land use if they are
spatially complementary—that is, falling within a walking distance
of the dominant land use. These concepts are at the core of the tools
developed here.
Street connectivity and parcel size are two other factors used in
PPI to measure potential demand for pedestrian travel. The use of
these variables is important, but they need refinement to capture
more accurately the characteristics of environments with latent
demand for pedestrian travel. In older urban areas—typically fully
developed before the Second World War—small street blocks and
parcels in mixed-use areas are indeed associated with complete non-
motorized infrastructure and some level of transit service, and they
usually command medium to high actual demand for pedestrian
travel. Yet these same relationships cannot be assumed to hold in
newer, suburban areas. Previous research showed that low actual
and high latent demand exists in newer areas that have large and
very large street blocks (1, 2). This is because dense and mixed
development of apartments, retail facilities, offices, or schools built
in the postwar period has taken place in large parcels that are in turn
served by large street blocks. In contrast, small parcels and small
street blocks in areas developed after the Second World War corre-
spond to low-density or even very low-density single-family zones
or subdivisions that are unlikely to generate substantial pedestrian
travel demand, actual or latent. Therefore, relating small parcels and
street blocks to potential pedestrian travel demand does not capture
the appropriate land use conditions in postwar development.
Previous research and tools have made important contributions to
estimating and locating pedestrian demand. The tools developed and
being used in Portland are especially ambitious. Critique of these
tools is meant not to belittle this work but to build and improve on
it to construct pedestrian location identifier tools (PLIs). PLIs are
especially designed to identify suburban, postwar areas that have
land use characteristics likely to create substantial latent demand for
pedestrian travel.
PEDESTRIAN LOCATION IDENTIFICATION TOOLS
Two tools were developed to identify locations with latent demand
for pedestrian travel. The two tools use different sets of databases.
PLI_1 relies on readily available census data, geographic informa-
tion system (GIS) software, and aerial photographs. PLI_2 uses
parcel-level data with GIS software. PLI_1 has a wide range of ap-
plications because the data are easily acquired. PLI_2 requires data-
bases that only metropolitan planning organizations, large cities,
and some counties and states may have (23, 24). The reliance on dif-
ferent databases affects the methods used to identify locations with
latent demand for pedestrian travel. PLI_1 is a manual method be-
cause of its dependence on the individual analyst’s judgment in com-
paring the data from the census with aerial photographs to delineate
clusters. PLI_2, on the other hand, is an automated method, because
the criteria for delineating the pedestrian locations are explicitly
defined and integrated into GIS-driven routines.
The tools are similar on several accounts. First, they rely on spa-
tial units of data that are small enough to capture the actual attrib-
utes of land development. Second, the same concepts structure both
tools to identify areas with potential for pedestrian travel. The fol-
lowing combination of generator and attractor land uses is selected,
between which relatively high volumes of pedestrian volumes are
likely:
•
Identification of a dominant land use, which represents a suffi-
cient market (generator) of potential pedestrians. The tools focus on
residential land use with residential density housing types serving
as measures of the dominant land use. PLI_2 can easily be modified
to focus on other types of dominant land uses.
•
Identification of functionally complementary land uses. The
tools focus on retail facilities and schools.
•
Definition of the spatial complementarity of land uses. The dis-
tance between complementary land uses is short enough that people
can realistically choose to walk instead of drive. A walking shed of
0.8 km in radius is used to define the maximum straight-line distance
between trip origin and destination.
Finally, use of the tools involves three basic steps. First, the
denser residential parts of suburban areas are highlighted as gener-
ators of pedestrian travel. Second, nearby retail areas are identified
as attractors. Third, areas characterized by dense residential and
retail development are delineated into clusters or zones with pedes-
Moudon et al. Paper No. 02-3748 97
trian travel potential. Figures 1 and 2 display potential pedestrian
zones by using PLI_1 and PLI_2, respectively, in Cluster A, the
same area of Puget Sound’s South King County.
Below are summaries of the specific protocols followed in each
of the tools. The methodological steps together with an illustrative
application to a specific area can be reviewed on the Internet
(www.wsdot.wa.gov/TA/PA andl/Bike_Ped/Guide_Reg.htm, www.
urbanformlab.net). Complete manuals to assist in the use of the tools
are also available (25).
PLI_1
The PLI_1 method uses U.S. population census data at the census
block level—the smallest spatial unit of data available—and aerial
photos covering the same extent—at a scale of approximately 1 =
24,000. The census data are used to identify areas in relatively dense
residential development (dominant land use and travel generator).
The aerial photographs serve, first, to show where dense residential
development takes places on the ground and, second, to visually
scout for the types of complementary land uses that are nearby areas
of concentrated residential development.
Queries of census data are used to identify blocks with population
densities above certain thresholds and blocks with high percentages
of denser housing types, specifically non-single-family housing and
0 1 mile
FIGURE 1 PLI_1. Census blocks delineating
clusters with potential demand for pedestrian
travel. Area depicted is city of Kent in South
King County, Washington. Light gray: Query 1
only (>10 people per acre); medium gray:
Query 2 (50% non-single-family) and in some
cases includes results from Query 1; dark
gray: Query 3 (apartments with >10 dwell-
ings per building) and in some cases
includes results from Queries 1 and 2.
multifamily units in structures with 10 or more units. Queries by
housing type are necessary because concentrations of dense hous-
ing may be located on very large blocks, resulting in low overall
densities.
Census blocks highlighted in GIS as a result of these queries rep-
resent a first attempt at identifying potential clusters. Spatial displays
show groups of identified blocks distributed throughout the area
being studied. The analyst eliminates single blocks and groups of
blocks that have low total populations. A threshold of 1,400 people
per clustered groups of blocks was used for applications in the Puget
Sound region. This was based on existing development patterns and
pedestrian studies in suburban areas in the region (1, 2, 18).
Areas defined by the highlighted census blocks are then analyzed
from aerial photographs. The photographs allow the analyst to estab-
lish the precise location of multifamily development shown in the
census blocks and to identify retail development, schools, or office
development (none of which are available in the census data) that
may lie close to the dense residential development. The analyst uses
the photographs to delineate areas with multifamily housing, retail,
and other selected uses in close proximity. The delineation process
makes it possible to isolate areas of concentrated development and
mixed uses from adjacent low-density development.
The spatial display of clusters delineated on the photographs
reveals both small and large areas of concentrated and mixed-use
development. A template defining walking sheds is then overlaid on
the potential clusters. A template representing a 0.8-km radius circle
[an area of 200 ha (500 acres)] was used in the Puget Sound work.
Some clusters of concentrated development will be 0.8 km in radius
or smaller, forming a potentially walkable area. Others may involve
98 Paper No. 02-3748 Transportation Research Record 1818
areas that are much larger than 200 ha. These will require further
analysis of the distribution of retail and residential land uses to
determine likely walking patterns within a 0.8-km radius. To do so,
the template needs to be centered on principal roadway intersections
with the most intense retail development around them as potential
neighborhood centers.
Census blocks are then reselected that best match the potential
cluster areas delineated by using the photographs and correspond-
ing to the 0.8-km distance from the central corner. The analyst will
find that census blocks often cover areas that are larger than the
actual clusters shown in the photographs and that stretch beyond the
walking shed. This is unavoidable because census blocks are defined
by street blocks, and many suburban street blocks are extremely
large. If large census blocks include areas of vacant land or sporadic
single-family development (they often do), then population densi-
ties obtained from the census data will be lower than the actual clus-
ter’s density. In a final step, the analyst will derive basic statistics
for each cluster from the census data. An example of such statistics
is shown in Table 1 for Cluster A, depicted in Figures 1 and 2.
PLI_2
PLI_2 uses parcel data to identify clusters. Parcel data allow for an
analysis of land use patterns at a very fine scale. They also permit the
use of GIS routines that specify the criteria used for defining pedes-
trian clusters. Conceptually, PLI_2 is similar to PLI_1. It defines
clusters by identifying the dominant land use and the land uses that
are functional complements, that is, land uses likely to generate
travel between them. It also tests whether these uses are concentrated
in areas that are small enough to be traversed by walking. This is the
test of whether functional complements are also spatial comple-
ments. If land uses are both functionally complementary and con-
centrated in small areas, they are defined as clusters that are likely to
have latent demand for pedestrian travel.
The first step in PLI_2 is to select out parcels with dominant and
complementary land uses from the entire set of parcels. Because
data are available for all parcels regardless of the type of land use on
the parcel, PLI_2 can collapse the identification of dominant and
complementary land uses into one step. In the Puget Sound work,
parcels were selected with residential development above 25 dwell-
ings per ha (10 units per acre) (generators), with retail (attractors),
and with schools (also attractors). The selection of this set of uses is
designed to identify mixed-use, residential neighborhoods as poten-
tial pedestrian zones. Additional uses, such as transit stations, parks,
and recreational facilities, can be included. Furthermore, the method
is suitable for different mixes of functionally complementary land
uses that may be found in employment centers. In either case, uses
not selected for cluster delineation are reintroduced at the end of
the delineation process to ensure a full accounting of parcels within
defined clusters.
The next steps test whether these selected land uses are within spa-
tially concentrated areas, and, if so, define them as clusters. PLI_2
uses a raster data model for this purpose. This model translates the
parcel polygons into grid cells to facilitate spatial analysis opera-
tions. Grid cells can take different specified values. In the PLI_2
raster model, cells in the same land use have the same value and rep-
resent patches as spatially contiguous areas of land in the same use.
A patch may correspond to one or many parcels. The raster model
thus displays dominant and complementary land uses as patches
of homogeneous land uses.
0
1 m
ile
FIGURE 2 PLI_2. Patches delineating
clusters with potential demand for pedestrian
travel (area depicted is same as in Figure 1).
Light and medium gray: residential; dark gray:
retail; black: schools.
To test whether the patches of selected land uses are spatially
complementary, a buffer is established around each patch, extend-
ing it outward by a specified distance. A 120-m (394-ft) buffer was
used in the Puget Sound work. Where buffers from one patch inter-
sect buffers from another, the patches are defined as being spatial
complements. They are within a specified distance of each other and
are defined as belonging to the same cluster. Note that buffers will
include not just the selected land uses but also vacant land, office
uses, and streets and roads. Other technical steps refine the delin-
eation of potential clusters. At the end of the process, areas are delin-
eated connecting many patches into the same cluster. A final step is
to recapture parcel data attributes. Some of these data are shown in
Table 2 for the example of Cluster A and can be used to analyze the
size, mix, and intensity of resultant clusters.
Moudon et al. Paper No. 02-3748 99
CONCLUSIONS
Both PLI_1 and PLI_2 enable the analyst to differentiate between
suburban areas that do and do not have potential for pedestrian travel.
They add to current practice, but they also require further research.
On the plus side, the tools offer the following improvements:
•
By considering combinations of land uses that are generators
and attractors of pedestrian travel, they capture the characteristics of
land use mixes that have the highest potential for substantial volumes
of pedestrian trips.
•
By using small spatial units of land use data, they adequately
capture the characteristics of development on the ground. At the
same time, because the small spatial units of data are available for
areas of large extent, the tools can be readily applied to today’s
spread-out cities and urbanized regions.
•
The small spatial units of data allow a precise and accurate
measurement of the land use characteristics of the small areas that
correspond to short walking distances.
•
By focusing on multifamily as the dominant land use, the
methods bring attention to suburban areas that have been neglected
in the past as locations with untapped potential for pedestrian travel.
At the same time, the tools will need to be applied to areas where
employment land uses dominate to complete the identification of
areas that deserve special consideration for latent pedestrian travel
demand.
It is worth mentioning that the tools have applications beyond
pedestrian travel. In general, they serve to identify areas of compact
development in suburban areas. Compact development is of increas-
ing interest to planners interested in smart growth as a form of land
development that is infrastructure efficient—infrastructure includ-
ing not only transportation systems but also water and utility sys-
tems. Compact development represents areas where special con-
sideration is needed for public service delivery—such as fire and
Tract No. Block Pop. Acres
Pop.
Density
Total
Housing
Units
1 Unit per
Structure
10+ Units
per
Structure
# Non-
Single
Family HU
% Non-
Single
Family
HU
Density
% Non-
White
%
<18 yrs
029404 212 1140 77.6
14.7 431.0 79 183 352.0 81.7 5.6 19.1 38.3
029404 208 73 5.9
12.3 21.0 20 0 1.0 4.8 3.5 11.0 32.9
029404 304 50 4.9
10.1 17.0 17 0 0.0 0.0 3.4 0.0 26.0
029404 303 82 4.2
19.5 21.0 21 0 0.0 0.0 5.0 19.5 36.6
029404 301B 229 11.9
19.3 75.0 45 26 30.0 40.0 6.3 14.0 30.6
029404 301A 994 45.2
22.0 515.0 11 360 504.0 97.9 11.4 9.5 23.5
029404 306 97 6.9
14.0 31.0 30 0 1.0 3.2 4.5 15.5 32.0
029404 307 95 8.6
11.0 32.0 32 0 0.0 0.0 3.7 6.3 26.3
029404 309 74 6.7
11.1 24.0 24 0 0.0 0.0 3.6 18.9 29.7
029404 310B 105 3.5
30.4 35.0 17 17 18.0 51.4 10.1 3.8 30.5
029404 213 83 17.3
4.8 52.0 2 5 50.0 96.2 3.0 6.0 8.4
029404 310A 54 3.5
15.6 16.0 16 0 0.0 0.0 4.6 38.9 33.3
029404 312 271 32.6
8.3 120.0 16 34 104.0 86.7 3.7 16.6 28.4
029404 311 47 3.7
12.7 23.0 19 0 4.0 17.4 6.2 4.3 29.8
029404 302 140 5.7
24.6 92.0 0 74 92.0 100.0 16.2 11.4 15.7
029404 214 47 5.4
8.6 28.0 0 27 28.0 100.0 5.2 10.6 14.9
029501 201 211 114.9
1.8 116.0 54 61 62.0 53.4 1.0 0.5 22.3
029501 901 858 119.3
7.2 422.0 39 322 383.0 90.8 3.5 14.3 26.3
Totals 4650 477.89
13.06 2071 442 1109 1629 43.34 5.29 11.59 25
.56
TABLE 1 Data Derived from Census Block Analysis for Cluster A
Attributes Measures Measures (metric)
Total Area 524 a 210 ha
Mean Block Size 48 a
19 ha
Total Lots 266
Mean Lot Size 1.7 a 0.
7 ha
Total Buildings 653
Total Housing Units 4232
Housing Density Gross 8.1/acr
20/ha
Commercial Floor Area 1.4 M. sq. ft. 137k sq. m.
Commercial FAR 0.18
Population 9321
Population Density 17.8 p/a 44.
5 p/ha
TABLE 2 Data Derived from Parcel Database for Cluster A
police. In this sense, the tools can be used in a number of public pol-
icy situations affected by the distribution and concentration of pop-
ulation and activities. Both tools allow planners to quantify the
attributes of compact development. PLI_2 has the advantage of ob-
jectivizing the delineation of clusters, a task that needs to be done
manually in PLI_1. PLI_2 also allows the planner to access a larger
set of data attributes that define land development than is available
in census data at the block level.
The tools demand further research to match the added precision in
measuring pedestrian-supportive land uses with added precision in
measuring the pedestrian friendliness of the transportation facilities.
The suburban areas identified typically are deficient in pedestrian
infrastructure, including low levels of connectivity and accessibility,
with large street blocks, incomplete sidewalk systems, large at-grade
parking lots, fenced-in development, and so forth. As a result, con-
siderable investment will be required to retrofit suburban clusters
with the safe and comfortable walking conditions that will support
substantial increases in pedestrian travel volumes. The Puget Sound
region has many of these places, and other regions are likely to as
well. Because of this, it will be necessary to go beyond identifying
areas with high latent pedestrian demand and prioritizing them for
investment above areas with low latent demand. Prioritization will
also need to take place between clusters to yield the highest benefits.
Portland’s PDI addresses this issue by measuring the characteris-
tics of transportation facilities that do or do not support pedestrian
travel. As part of this research, a third tool, the pedestrian infra-
structure prioritization decision system tool (PIP), was sketched out
to guide further selection of clusters where investments will gener-
ate the highest potential benefits for pedestrian travel (25). PIP dif-
fers from PDI in that it integrates both land use and transportation
facility considerations.
Finally, and importantly, the methods proposed attempt to ad-
vance both the concepts and the measures used to capture the char-
acteristics of pedestrian-supportive land use and development pat-
terns. They attempt to build on previous efforts. Like those efforts,
however, they remain mostly theoretical and untested. Field data are
needed to further the understanding of the theoretical constructs dis-
cussed and to yield tools that can actually quantify the demand for
pedestrian travel associated with given land use and development
patterns, as follows:
•
The concept of a dominant land use as the generator of pedes-
trian travel deserves further attention. Although future measurements
of land use mix that are conducive to pedestrian travel should not
assume covariance between residential and employment land use in-
tensity, the respective demands for such travel in concentrations that
are primarily residential versus primarily employment land uses need
investigation.
•
Given a dominant land use, it will be necessary to know the
effect of such single variables on walking volumes as increased resi-
dential or employment density and the presence of specific attractors
or generators such as schools and parks.
•
Similarly, the concept of functionally complementary land uses
needs to be better operationalized. What land uses are best linked by
pedestrian travel? for whom?
•
The spatial complementarity of land uses needs testing as well;
particularly, the 0.8-km walking shed must be validated as applied
to different land uses, populations, and topographical conditions.
•
Given pedestrian-supportive land use and development patterns,
the effect of different levels of service in transportation infrastructure
needs to be better understood.
100 Paper No. 02-3748 Transportation Research Record 1818
Such data gathering will require a substantial funding commit-
ment. It is essential, however, in order to quantify the relationship
between land use and travel behavior and, specifically, to properly
estimate expected pedestrian volumes. Such estimates will yield the
best approaches to prioritizing future infrastructure investments.
Overall, this and other work suggest that some of the land use and
development patterns generated over the last half-century could sup-
port pedestrian travel. The relatively high proportion of short car
trips shown in transportation data further substantiates the possibil-
ity of substituting these with walking. Yet although some of the en-
vironmental conditions and trip characteristics for walking are not
entirely discouraging, improvement in the share of nonmotorized
travel is unlikely unless further changes are made in both public
policies and attitudes toward nonmotorized travel. The public must
be made aware of the real costs of the escalating number of motor-
ized trips and must want to reduce them. To this end, transportation
data must focus on transit riders, walkers, and bikers. Finally, the
share of public spending for infrastructure supporting pedestrian and
nonmotorized travel must be increased substantially.
ACKNOWLEDGMENT
The authors thank the six anonymous reviewers for their thoughtful
and constructive comments.
REFERENCES
1. Hess, P. M., A. V. Moudon, M. C. Snyder, and K. Stanilov. Site Design
and Pedestrian Travel. In Transportation Research Record: Journal of
the Transportation Research Board, No. 1674, TRB, National Research
Council, Washington, D.C., 1999, pp. 9–19.
2. Moudon, A. V., and P. M. Hess. Suburban Clusters: The Nucleation
of Multifamily Housing in Suburban Areas of the Central Puget
Sound. Journal of American Planning Association, Vol. 66, No. 3,
2000, pp. 243–264.
3. Pucher, J., and L. Dijkstra. Making Walking and Cycling Safer:
Lessons from Europe. Transportation Quarterly, Vol. 54, No. 3, 2000,
pp. 25–50.
4. Guidebook on Methods to Estimate Non-Motorized Travel: Supporting
Documentation. Case Study 9. FHWA, U.S. Department of Transporta-
tion, 1999.
5. Nationwide Personal Transportation Survey. FHWA, U.S. Department
of Transportation, 1990.
6. Municipal Strategies to Increase Pedestrian Travel. Washington State
Energy Office, Olympia, 1994.
7. Dunphy, R. T. Transit Trends. Urban Land, May 2001, pp. 79–83.
8. The Pedestrian Environment, Vol. 4A. Making the Land Use, Trans-
portation, Air Quality Connection (LUTRAQ). 1000 Friends of Oregon,
Portland, 1993.
9. Accommodating Bicycle and Pedestrian Travel: A Recommended
Approach. FHWA. 2000. www.fhwa.dot.gov/environment/bikeped/
design.html. Accessed July 10, 2001.
10. Bicycle and Pedestrian Data: Sources, Needs & Gaps. U.S. Bureau of
Transportation Statistics, U.S. Department of Transportation, 2000.
11. Antonakos, C. L. Nonmotor Travel in the 1990 Nationwide Personal
Transportation Survey (with Discussion). In Transportation Research
Record 1502, TRB, National Research Council, Washington, D.C.,
1995, pp. 75–82.
12. Holtzclaw, J. Using Residential Patterns and Transit to Decrease Auto
Dependence and Costs. National Resources Defense Council, San Fran-
cisco, Calif., 1994.
13. Newman, P. W. G., and J. R. Kenworthy. Cities and Automobile Depen-
dence: A Source Book. Ashgate Publishing Co., Brookfield, Vt., 1989.
14. Levine, J., I. Aseem, and T. Gwo-Wei. Innovation in Transportation
and Land Use as Expansion of Household Choice (in preparation).
15. Rutherford, G. S., J. M. Ishimaru, J. Chang, and E. McCormack. The
Transportation Impacts of Mixed-Use Neighborhoods. Washington
State Transportation Commission, Olympia, 1995.
16. Leinberger, C. B. Metropolitan Development Trends of the Late 1990s:
Social and Environmental Implications. In Land Use in America (L. H.
Diamond and P. F. Noonan, eds.), Island Press, Washington, D.C., 1996.
17. Moudon, A. V., P. M. Hess, M. C. Snyder, and K. Stanilov. Effects of
Site Design on Pedestrian Travel in Mixed-Use, Medium-Density Envi-
ronments. In Transportation Research Record 1578, TRB, National
Research Council, Washington, D.C., 1997, pp. 48–55.
18. Hess, P. M. Pedestrians, Networks, and Neighborhoods: A Study of
Walking and Mixed-Use, Medium-Density Development Patterns in the
Puget Sound Region. Ph.D. dissertation. University of Washington,
Seattle, 2001.
19. Oregon Bicycle and Pedestrian Plan. Bicycle and Pedestrian Program,
Oregon Department of Transportation, Salem, 1995.
20. Landis, B., and J. Toole. Using the Latent Demand Score Model to
Estimate Use. In Pro Bike Pro Walk 96: Forecasting the Future.
Bicycle Federation of America–Pedestrian Federation of America, 1996,
pp. 320–325.
Moudon et al. Paper No. 02-3748 101
21. Dixon, L. Adopting Corridor-Specific Performance Measures for Bicycle
and Pedestrian Level of Service. Transportation Planning, Traffic Engi-
neering Department, Gainesville, Fla., 1995.
22. Hess, P. M., A. V. Moudon, and M. G. Logsdon. Measuring Land Use
Patterns for Transportation Research. In Transportation Research Record:
Journal of the Transportation Research Board, No. 1780, TRB, National
Research Council, Washington, D.C., 2001, pp. 17–24.
23. Kollin, C., L. Warnecke, W. Lynday, and J. Beattle. Growth Surge:
Nationwide Survey Reveals GIS Soaring in Local Governments. Geo
Info Systems, Feb. 1998, pp. 24–30.
24. Monitoring Land Supply with Geographic Information Systems: Theory,
Practice, and Parcel-Based Approaches (A. V. Moudon and M. Hubner,
eds.), John Wiley and Sons, New York, 2000.
25. Moudon, A. V. Targeting Pedestrian Infrastructure Improvements: A
Methodology to Assist Providers in Identifying Suburban Locations with
Potential Increases in Pedestrian Travel. WA-RD 519.1. Washington
State Transportation Commission and U.S. Department of Transporta-
tion, 2001. www.depts.washington.edu/transnow/Research/Reports/
TNW2001-05.pdf.
Publication of this paper sponsored by Committee on Pedestrians.
