MODELING

WALKABILITY
Automating analysis so it
is easily repeated
By Arjun Rattan, Anthony Campese, and Chris Eden,
Regional Municipality of Halton, Ontario, Canada

The health and planning departments in one of Canada’s regions have  This model assigns total residential population to an analysis area.
joined forces to analyze just how accessible stores and services in
neighborhoods are by foot or bike. Walkable neighborhoods can also
reduce pollution from greenhouse gas emissions generated by vehicles. Quantifying Walkability
The GIS Services section within Halton Region’s Planning Services For this project, three factors—density, diversity, and design—were
Division, in collaboration with Halton Region’s Health Department calculated for use in measuring walkability. Density was quanti-
in Ontario, Canada, recently completed an analysis workflow to de- fied as residential population density and job density. Diversity was
termine the walkability of the region. [The region covers more than quantified as the percentage of residents within walking distance
232,000 acres, including a 25-kilometer frontage onto Lake Ontario, of defined diverse uses (DUs) that provide a measure of mixed-used
and has an approximate population of 467,200.] development. Design was quantified as trail availability per 1,000
ArcGIS 9.3.1 geocoding tools, along with the Network residents, bicycle path availability per 100 residents, and the number
Analyst 9.3 extension, were key technologies used to determine of intersections per square kilometer.
Halton’s walkable areas. ArcGIS ModelBuilder 9.3 enabled automa- These factors were equally important when assessing the walkabil-
tion of the workflow, making the process both quick and easily re- ity of an area. From a GIS standpoint, calculations relating to density
peated without requiring extensive technical GIS knowledge. and diversity were relatively simple and enabled by ArcGIS technology.

 These screen captures illustrate the difference between service areas for a post office created using Euclidean distance buffering (left)
and street network buffering (right). The residential parcels northwest of post offices are not classified as walkable when Network Analyst is
used to create the buffer. In this realistic scenario, a water body impedes walkability of these post office locations. Note: Sample data used
to protect privacy of residents.

30 au Winter 2012 esri.com

 Feature

Five analysis areas (AAs) in the Halton Region were specified for
the project:
• Dissemination areas (small areas composed of one or more neigh-
boring blocks with a population of 400 to 700 persons)
• Census tracts
• Neighborhoods
• Municipalities
• The Halton Region as a whole
Dissemination areas were the smallest AAs because they showed
an acceptable level of detail while still allowing comparisons to
census data.
Halton Region’s Health Department defined three parameters  A model for calculating the total residential population of an
needed to determine and quantify diversity: analysis area
• AAs for which walkability needs to be determined
• DUs needed by residents
• The distances to/from DUs (in meters) considered walkable
distances
Table 1 lists the DUs used for the project, along with the walkable What Is Walkability?
distance criteria to the DU location. These distances represent 5-
and 10-minute walking times (except for elementary schools). Walkability is a measure of the effectiveness of
community design in promoting walking and bicy-
Gathering and Preparing Data for Analysis cling as alternatives to driving cars to reach shop-
Datasets required to conduct the walkability analysis were AAs as
ping, schools, and other common destinations. The
polygons, streets as lines, residential locations with population infor-
mation as polygons and points, and DU locations as points. The AAs Centers for Disease Control and Prevention (CDC),
and streets layers were available from the region’s spatial database. the World Health Organization (WHO), and other
Residence locations were extracted from the Ontario Municipal health organizations advocate increasing the walk-
Property Assessment Corporation (MPAC) parcel layer. The MPAC
ability of communities to promote fitness, combat
parcel, street, census tract, municipality, and region layers were cur-
rent as of 2011. The dissemination areas fi le was from 2006, and the obesity, and enhance sustainability.
neighborhoods fi le was from 2009.
The MPAC parcel property code (PROP_CODE) and structure
code (STRUCTURE_) fields were used to identify residential parcels.
These parcels were exported and formed the residential parcel loca- number of units on the parcel to get the total population for the
tion polygon layer. A persons per unit (PPU) attribute field, added to parcel. Th is value, stored in the UNIT_POP field, described the total
the residential parcel layer, described the average number of people population of a residential parcel. Calculating population using the
associated with a particular residential dwelling type (e.g., PPU for PPU field enabled estimating current population on a per parcel
single-family dwelling units was 3.08). basis. Moreover, census population information was outdated; the
PPU values were calculated by Halton Region’s Planning Services. most recent census information was 2006.
For parcels with multiple units, parcel roll numbers were used with A copy of the residential parcels polygon layer was exported as
the MPAC Municipal Connect online database to verify the number a point layer using the Feature To Point tool (ArcToolbox > Data
of residential units on a parcel. The PPU field was multiplied by the Management > Features). Both residential parcel polygon and point
layers were used later in the analysis workflow.

Diverse Use (DU) Walking Distance Criteria Geocoding DU Addresses

(i.e., distance from residence to DU) Of the DU data items listed in Table 1, only transit stops were avail-
able as a shapefi le. Most DU location addresses, available from a
Transit stops ≤ 400 meters Microsoft Access database, were gathered as part of the 2010 Halton
Grocery stores ≤ 800 meters Region Places of Employment Survey. The required DUs were iden-
tified (and exported as Microsoft Excel tables) by North American
Fresh food places ≤ 800 meters Industry Classification System (NAICS) code field. (NAICS identifies
business and service types.) Addresses for postal service locations
Convenience stores ≤ 400 meters
(not available from the employment survey) were gathered from the
Elementary schools ≤ 1,500 meters Canada Post website. Education Services addresses were gathered
by contacting school boards in the region, in addition to the address-
 Table 1: A sample of DUs used in the project along with walking es of private and public higher education institutions within
distance criteria

esri.com Winter 2012 au 31

 Ñ Model for calculating
service areas for diverse uses
and exporting it as a polygon

Halton. Similarly, Fresh Food Places address data was gathered by buffer around the DU was initially considered as a solution to create
contacting local organic and fresh food stores. the SAs. However, this method wasn’t realistic from a walkability
Having the DU locations available as a point shapefi le was critical standpoint because it didn’t account for physical barriers such as
to the walkability analysis workflow. ArcGIS 9.3.1’s geocoding tools water bodies, railway lines, buildings, and other obstructions.
converted DU addresses into spatial points. A new geocoding ad- Network Analyst 9.3 tools were used in ModelBuilder to design an
dress locator was created using an address point shapefi le. The Zone algorithm that calculated the SA for a DU, using the Halton street
field in the address locator enabled identification of street addresses network. Th is SA model represented a more realistic visualization
with the same street name that are located in different municipali- of walkable access to a DU by using the street network rather than
ties (for example, 543 King Street exists in both Oakville and Halton Euclidean distance.
Hills). A high minimum match score minimized the chance of a DU In ArcCatalog, the Halton streets layer (Streets_Ctrline.shp) was
address being matched to the wrong address point. used to create a new network dataset for network analysis. The dis-
The automated geocoding process achieved a success rate of more tance in meters was the only cost associated with the network data-
than 90 percent for all DU addresses geocoded. The Pick Address set, because no network analysis involving travel time was required.
From Map tool in the Geocoding Review/Rematch Addresses func- After the SA is calculated, it is exported as a polygon layer. The Select
tion helped bring the geocoding match rate up to 100 percent. Th is Facility parameter specifies the DU for which SA is to be calculated.
tool was used to pick points on the map that were verified as loca- Created SA polygons were merged and trimmed to 30 meters. This
tions of unmatched addresses. distance was determined to be the best distance for ensuring that
only residential parcels identified as walkable to the DU were selected.
Determining Diversity
Having gathered and prepared the data required for analysis, the Select Residential Parcels That Intersect the SA
next steps involved determining the percentage of residents who for Each DU
live a walkable distance from DUs. Analysis involved repeating This step involved selecting residential parcels within walkable dis-
workflow steps for different input parameters such as AAs and DUs. tance from a specific DU. The selected residential parcels were con-
ModelBuilder 9.3 was used to automate this process. The workflow verted to residential parcel points to prevent walkable parcels that
steps for this analysis included fall on AA borders from being included in multiple AAs in the next
• Calculate and assign the total residential population for each AA. step.
• Calculate a service area (SA) for a DU.
• Select residential parcels that intersect the SA for each DU. Calculate the Percentage of Residents in Each AA
• Calculate the percentage of residents in each AA with walkable with Walkable Access to Specified DU
access to a specified DU. Listing 1 illustrates how the percentage of residents with walkable
• Classify and visualize walkability. access to a specific DUis calculated. Th is step also assigns the calcu-
lated proportion to the AA shapefi le.
Calculate and Assign the Total Residential Dim x as Double
Population for Each AA If ([MAX _ TotRes] > 0) Then
The total residential population was calculated and assigned as the x = [SUM _ UNIT _ P] / [MAX _ TotRes] * 100
sum of the UNIT_POP field for all residential parcel points that were Else
within the specified AA. In assigning total residential population to X = 0
an AA, residential parcel points were used instead of residential par- End If
cels to prevent parcels that fall on the border of multiple AAs from
 Listing 1: Percentage of population in walkable residences
being included in the total population calculation of multiple AAs.
The Visual Basic for Applications (VBA) code in Listing 1 was used
Calculate Service Area for a DU in the Calculate Field tool to calculate the proportion of residents
The SA for a DU was based on the walking distance criteria in Table 1 (PR16DUPost field) that live a walkable distance from a post office. A
and represented all the locations that have walkable access to the conditional statement is required to prevent illegal division by 0 (in
specified DU. Using straight-line Euclidean distance to create a the case that total residential population of an AA is zero).

32 au Winter 2012 esri.com

 Feature

Classify and Visualize Walkability

The results of the walkability analysis were used to construct a da-
tabase that contained walkability calculations for density, diversity,
and design for each dissemination area. This database was used to
summarize results for census tracts, neighborhoods, municipalities,
and the Halton Region. Choropleth map layouts for various DUs for
different AAs were produced to visualize walkability to DUs. The  Model for selecting residential parcels that are in walking
maps displayed AAs classified by the walkability percentage field. The distance to a specified diverse use
areas in red represent the areas with low walkability to elementary
schools. A separate walkability map for each DU enables the planning
process to focus on services that are most in need, for a given AA. GIS technologies that make a realistic determination and visualiza-
tion of walkability feasiblity. ModelBuilder, in particular, enables pro-
Project Repeatability ject repeatability while minimizing project completion timelines.
The walkability project took approximately three months to com-
plete. A key goal for this project was repeatability. The analysis About the Authors
workflow should easily be repeated with updated data and criteria. Arjun Rattan is a GIS planning assistant at Halton Region’s
Creating a custom ArcGIS toolbox that houses custom ModelBuilder Planning Services Division. Before joining Halton, he earned a bach-
models enables the project to be repeated in significantly shorter time elor’s degree in urban studies and art history from the University of
periods by users who don’t have extensive technical GIS knowledge. Toronto and completed a two-year diploma in geomatics from Sir
The entire walkability analysis workflow for the diversity factor Sandford Fleming College in Lindsay, Ontario.
(excluding data gathering/preparation) can be repeated by running
the four models in the custom Walkability Tools toolbox developed Anthony Campese is a GIS analyst with the Legislative & Planning
for this project. Running the models in batch mode lets models use Services Department who is responsible for maintaining the GIS
multiple inputs simultaneously and saves both time and effort. functions in the Planning Services Division and acts as a liaison with
IT staff. He has worked in the public sector for 12 years. He is also
Conclusion an instructor in real property administration and public adminis-
An area’s walkability is not a single quantifiable measurement. tration—municipal programs at Seneca College. He is a graduate of
Walkability, in this study, was measured using calculations related Fleming College’s GIS Applications Specialist program and holds a
to an area’s density, diversity, and design. These three factors provide degree in urban geography from Brock University.
valuable information for assessing an area’s walkability. Examining
the database and walkability maps of the various DUs for different Chris Eden is a GIS data technician with the Legislative & Planning
AAs provides a context for determining whether an area needs more Services Department. He is responsible for maintaining the plan-
social marketing to encourage walking (if walkability is high) or better ning data and field collection of data. Eden has been working in the
design to increase access (if walkability is low). ArcGIS 9.3.1 geocoding public sector for eight years. He is a graduate of Ryerson University’s
tools, ModelBuilder 9.3, and the Network Analyst 9.3 extension are key Geographic Analysis program.

 Intersecting residential
parcel points for Home,
Furniture, Furnishings, and
Equipment Stores service areas

esri.com Winter 2012 au 33