A Web-based Accessibility Toolkit for Transportation Planners

Howard SlavinAndres RabinowiczGiovanni Flammia

Caliper CorporationOctober 2013

Project Overview

• FHWA-sponsored project

• Competitive response to a Broad Agency Announcement (BAA) for projects that could be transformational for transportation planning

• The BAA primary goal was the development of tools and techniques that support state and local capacity building

• A secondary goal was to develop tools and techniques that support Tribal planning, public involvement, environmental justice, and visualization in planning

The Toolkit Approach

• Provide a free web-based tool that can be used by planners and citizens

• Deploy simple, but powerful analytical tools

• Exploit advances in GIS, web software, and data availability

• Focus on the ultimate goal of transportation initiatives

• Provide a rich and valid alternative to travel demand models

Accessibility Concept

• Conventionally traffic or mobility measures are used in place of accessibility measures, but “Accessibility is the ultimate goal of most transportation and so is the best approach to use” Todd Litman 2003

• Accessibility refers to the possibilities of traveling to destination opportunities and the level of service associated with a wide range of travel options

• Can be assessed for both person travel and freight movements and for all modes, not just vehicular modes

• Increasing travel options or their quality and performance increases accessibility

The 5-Cs of Effective Accessibility Measures

• As identified by Kevin Krizek and David Levinson

• Cumulative

• Comparable

• Clear

• Comprehensive

• Calculable

Accessibility Measures

• Develop easy to understand measures• Examples of used, but complex measures

– Gravity based (Hansen)

– Utility Based •

• Instead, we use time as the measure

The Concept is Old, Not New!

• Extensive literature extolling the virtues of the approach

• Numerous, but limited examples of U.S. application

• A key aspect of British planning where it is part of the fabric of “evidence-based” planning

• Can be used to examine all parts of a study region and to compare them and perform evaluation of plan components using consistent and low cost methods

Accessibility Measure Examples

• Travel time by mode

• Average travel time by mode

• Count of jobs, people within a travel time band

• Travel time savings with and without a project or improvement

Toolkit Transportation Modes

• Car

• Transit

• Bicycle

• Pedestrian

Data Requirements

Layers• Routable Street line

layers• Transit Route Systems• Demographic data

files• Employment Data• Landmark point

databases

Potential Data Sources• TIGER Streets and

Boundaries• GTFS Transit DATA• Census DATA• MPO Planning Model

Data• Local GIS data• User-supplied data• Further needs

Data Issues

• Routable streets not freely available • Planning model road networks are

available but generally insufficient in the level of detail and may be insufficient in many other ways

• Walk networks and bike networks are generally not prepared or prepared carefully by modelers

• Transit poses particular problems despite the availability of GTFS

Using GTFS to Measure Transit Accessibility

• Acronym for General Transit Feed Specification originally called the Google Transit Feed Specification

• Common format for transit schedules and related information

• Used by transit operators to submit data to Google

• A large number of transit agencies have public feeds

GTFS Pros

• The GTFS format and the GTFS data exchange have made transit data more widely available

• The GTFS format forces transit agencies to think about geographic locations by requiring a longitude and latitude for physical stops.

More GTFS Pros, But some Cons

• Makes transit data more widely available– GTFS data is available for 595 Agencies (May

2013)• Provides geographic coordinates for

transit stops• Route geography is often poor or missing• Does not include a street reference layer• To use one, conflation is required.

Comparison of Route Alignments and GTFS Shapes at Dupont Circle

Comparison of Route Alignments and GTFS shapes in Virginia

GTFS Conflation

• Requires a geographically accurate, routable street layer

• Even with good tools, some manual fix-up is probably inevitable

• Can be difficult even for experts

Examples of Conflated Transit Routes

Computational Approach

• TransCAD accessibility computation engine

• Custom web interface

• Generation of Comparative Maps and Reports

• Several technical challenges

Computational Elements

• Modal travel time skims between all intersections

• In our North Carolina example, there are 76,000 road nodes leading to 5 Billion path calculations.

• Contour generation for time bands

• Polygon overlay to obtain demographics

Network Bands Challenges

• Efficiency building band isochrones• Accuracy

– Account for major water bodies– Account for special links such as bridges,

tunnels and ferries– Account for sparse networks (national

parks, difficult topography) and out of analysis region areas

• Efficiency overlaying isochrones with demographic layers for data reporting

Starting Point: Band Area 6.9 sq. mi)

Without Exclusion Areas (5, 10, 15 min)

With Exclusion Areas

Area Difference

From (min) To (min) Area without Exclusion (sq. mi)

Area with Exclusion (sq. mi)

0 5 13.11 10.23

5 10 55.80 46.57

10 15 115.91 102.38

Bridges and Tunnels

Bridges and Tunnels

Treatment of Long Links

MapPoint 5 min Band

MapPoint 15 min Band

MapPoint Band vs. Path Inconsistency

Planning Networks vs. Accurate GIS

Planning Network Dense GIS Layer

Triangle NC Case Study

Base Scenario 2040

Links 92,174 93,708

Nodes 76,863 77,474

Routes 195 287

Stops 4,433 4,904

Transit Pathfinding Nuances

• “Best” transit paths include walking and possibly other modes

• In dense areas, there may be many alternative transit paths and these may overlap considerably

• Transit travelers preferences for routes vary with differing willingness to walk or wait or transfer

• Habitual users know about reliability, crowding, and other route choice factors

• Simple, schedule-based routing is not sufficient for measuring accessibility

• Our accessibility calculator can handle many route choice mechanisms

Transit Bands: Matrices

Data Requirements

• Road Layer (GIS, Transportation Model Networks)

• Transit Route Systems (Transportation Models, GTFS)

• Landmark database

• TAZ or similar polygon layer with demographic attributes to report

• Water areas and major exclusion areas

Process Description

Select a starting or Destination Point

Calculate Paths

Build Contours

Aggregate data by Time band

Report data

Functional Requirements: End User

Users

Compare

Travel Modes

Scenarios

Travel Times

Share

Print

Email

Download

Results

Functional Requirements: Administrator

Administrator

Add New Region

Add New Transportatio

n ScenarioNew or

Updated Demographi

c

New GIS Layers

Scale website

Application Architecture

Client Side API

• HTML single-page application – no plugins

• Javascript Library – Toolkit.app– Toolkit.Share– Toolkit.Gallery– Toolkit.Compare– Toolkit.Scenarios

•

Scenario Management

– Javascript, self documentedtoolkit.scenarios[“Base"] = {…overlay_fields: [["HHs", "Households"],

["TOT_POP", "Population"], ["TOT_EMP", "Jobs"]],

sizes: [5, 10, 15], min_fields: [["TIME", "Minutes"],

["LENGTH", "Miles"]], // List of available travel modes travel_modes: [["car", "Car"],

["walking", "Walking"], ["bike", "Bike"], ["bus", "Bus"]]

…}

– Hierarchy Structure (Project Scenario = Base Scenario + Differences) toolkit.scenarios[“Project"] = extend(toolkit.scenarios[“Base”], { min_fields: [ “TIME2040” …] })

Server Side

• Thin layer– ASP.NET Web Application

• Request broker, communicates with the GIS Web Server Farm

• Caches user request and responses • GIS Web Services from TransCAD for the

Web– Parallelisms via multiple instances of TransCAD– Each Instance highly multi-threaded– Drawing Engine– Geocoding, reverse geocoding– Accessibility calculator

Application Interface

Wake Memorial and Duke Hospital by Transit

Accessibility Demographic Measures

Comparison Side by Side (Transit vs Car)

Comparison Side by Side (car vs. transit)

Overlay Comparison

10 minute access band for a route

Quarter mile around all transit stops

Quarter mile transit stop accessibilityField Value (1/4

mile)Whole

RegionPercentage

Population 277,384 1,641,484 16%

Hispanic Origin 38,514 169,924 22%

White 129,845 1,000,210 13%

Black 88,460 370,809 24%

American Indian 1,294 8,984 14%

Asian 16,751 79,224 21%

Median Income 43,027 60,785

Avg. Per Capita Income 25,795 30,379

Median Family Income 59,037 73,080

Some Conclusions

• Very detailed data is needed for meaningful resolution

• A computationally intense process is therefore required

• Local data gathering and collection will typically be needed unless NAVTEQ® or alternative routable street database is available

• Unfortunately, a fair amount of work will be involved for any application

Possible Applications

• Travel Demand Model Support• Transit Planning• Bicycle and Pedestrian Planning• Smart Growth Analysis• Environmental Justice• Public Outreach• Performance Monitoring

Transit/Non-motorized Applications

• Market Analysis• Identification of Service Geography• Point to point level-of-service (LOS)

measures• Comparisons with auto LOS• User benefit mapping• Bike Lane/ sidewalk mapping• Non-motorized transit connectivity

EJ Applications

• Title VI Calculations• Identification of Under-served Transit

Markets• Distribution of Incremental User Benefits

by Income and Ethnicity

Software Availability

• The toolkit software is available now from Caliper

• One free copy is available to any U.S. State DOT, MPO, or public transit agency

• Data, data preparation, and implementation support are not included