Emerging Practices and Data Sources for Multimodal Transportation Planning, Design, and

Performance Monitoring

Borna Arabkhedri

A thesis

submitted in partial fulfillment of the

requirements for the degree of

Master of Science in Civil Engineering

University of Washington

2021

Committee:

Don MacKenzie

Xuegang Ban

Program Authorized to Offer Degree:

Civil and Environmental Engineering

©Copyright 2021

Borna Arabkhedri
 University of Washington

Abstract

Emerging Practices and Data Sources for Multimodal Transportation Planning, Design, and

Performance Monitoring

Borna Arabkhedri

Chair of the Supervisory Committee:

Don MacKenzie

Department of Civil and Environmental Engineering

An increasingly multimodal transportation network and the advent of new mobility solutions

means that jurisdictions should change their approaches in the planning, design, and

performance monitoring of transportation systems. The work is divided into two main parts.

The first section is an attempt to improve the understanding of effective multimodal

transportation design processes and performance-based decision making through conducting

a national-scale literature and document review. The conventional design approach of

prioritizing automobiles has led to problems such as the deterioration of the environment and

reduction in city quality of life. Therefore, multimodal design is becoming more widely

adopted by jurisdictions through designing for multiple travel modes rather than cars only.

However, as this is a relatively new field, the information available falls slightly short of the

demand and there is not an ultimate source of guidance for effective multimodal design

processes. This section aims to fill that gap by synthesizing "the state of the knowledge" via

systematically reviewing the available academic literature, as well as "the state of the
practice" by reviewing practical documents such as state guidebooks and design manuals. For

this study, multimodal will include: active transportation (e.g., pedestrians and bicycles),

freight, transit, and single-occupant vehicles. The work compares the current design

processes across various jurisdictions and draws conclusions on what are the best practices,

data sources, and performance measures for multimodal design.

The second section of the thesis focuses on a new data-gathering tool and analysis framework

that helps cities more effectively monitor the performance of micromobility services (i.e.,

shared e-scooters and e-bikes). As many of these services are nowadays dockless, the trips

can be ended almost anywhere and not necessarily in a parking hub, which leads to these

vehicles often getting mis-parked by the user, blocking sidewalks, or causing issues for

people with disabilities. We developed an application for the purpose of crowdsourcing

parking data from city residents to ensure that public servants get alarmed as soon as a

parking infraction gets reported and to help as a long-term data solution. This tool can also be

used for conducting data-driven parking audits of bikes and scooters, on a neighborhood or

even a city level. We used the app to conduct two parking audits in the cities of Portland and

Seattle. Summary statistics of these case studies are mentioned along with spatial analysis of

the data. Several statistical models were also developed to seek for any links between parking

violations and elements of the built environment and/or census tract socioeconomic factors

and demographics. One of the models for the City of Seattle shows that the number of bike

racks negatively correlates with the number of infractions on a census tract level. Even

though the results may be biased due to low sample size, the tool itself along with the

analysis piece can be used as a framework for other cities to conduct parking audits of

micromobility companies.
Acknowledgements
Completing this thesis would have not been possible without the help, guidance, and support

of numerous people from the University of Washington, my family, and my friends.

I would like to thank my advisor, Professor Don MacKenzie, for his tireless efforts in guiding

me throughout the various steps of the projects I was involved in and for supporting me in the

past two years. Your feedback sharpened my thinking and strengthened my work, and I am

grateful for having been able to work with you and being part of the Sustainable

Transportation Lab.

Furthermore, thank you to all my wonderful colleagues in the lab who were there for me

whenever I had questions and from whom I learnt a great deal.

Getting a master’s degree abroad and away from my family was not an effortless task, and I

would have not been here today without my family’s continual inspirational and emotional

support. I am grateful to my mother, father, and sister for their love and support.

Lastly, studying and making progress throughout the COVID-19 pandemic was not easy, but

great friends definitely made it much easier. I am grateful for all my close friends who

supported me in these months and helped make life a lot of fun outside of work and research.
Table of Contents
Acknowledgements .................................................................................................................... 5

1 Introduction and thesis overview ........................................................................................ 1

2 A review of multimodal design guidelines, processes, and performance measures ........... 3

2.1 Introduction ................................................................................................................. 3

2.2 Literature review ......................................................................................................... 4

2.2.1 Conventional approach to performance measurement ......................................... 6

2.2.2 A shift towards multimodal and reshaping communities .................................... 6

2.2.3 Definitions of performance measures, quality of service, and level of service ... 7

2.2.4 What is multimodal performance measurement? ................................................ 8

2.2.5 Review of multimodality in the Highway Capacity Manual (HCM) ................ 11

2.2.6 Multimodal Level of Service: old and recent methods ...................................... 17

2.2.7 Developing performance measures for project evaluation ................................ 21

2.2.8 Supply- vs. Demand-based performance measures ........................................... 32

2.2.9 Conclusions ........................................................................................................ 34

2.3 Review of best practices ............................................................................................ 36

2.3.1 Key terminology and antecedents to multimodal design ................................... 36

2.3.2 Informational interviews for identifying leaders in multimodal design ............ 39

 2.3.3 A review of jurisdictions’ available documents in multimodal, performance-

based, and/or practical design ........................................................................................... 44

2.3.4 Comparison of project development and design processes through key design

steps and elements ............................................................................................................ 52

2.3.5 Conclusions ........................................................................................................ 65

3 MisplacedWheels, A crowdsourcing tool for gathering micromobility parking infraction

data: the results of two parking audit case studies in Seattle and Portland.............................. 66

3.1 Introduction and background .................................................................................... 66

3.2 Benchmarking micromobility infraction reporting tools and results from interviews

with five jurisdictions .......................................................................................................... 69

3.3 Review of parking infraction rules in various cities ................................................. 73

3.4 App development ...................................................................................................... 76

3.5 Data audit conceptualization and sampling plan ....................................................... 77

3.5.1 Sampling method for Portland ........................................................................... 79

3.6 Data cleaning, post-processing, and summary statistics ........................................... 83

3.6.1 Data cleaning and post-processing..................................................................... 83

3.6.2 Summary statistics ............................................................................................. 86

3.7 Spatial analysis .......................................................................................................... 92

3.8 Data modeling and results ....................................................................................... 102

 3.9 Findings and conclusions ........................................................................................ 106

4 Conclusion ...................................................................................................................... 108

References .............................................................................................................................. 110

Appendices ............................................................................................................................. 120

Appendix A ........................................................................................................................ 121

List of Figures
Figure 1: Literature review guide .............................................................................................. 5

Figure 2: Classification of performance measures by mobility dimension (Elefteriadou et al.,

2012) ........................................................................................................................................ 27

Figure 3: Performance measures by modal perspective and mobility category (Dock et al.,

2017) ........................................................................................................................................ 27

Figure 4: Processes for developing performance-based measures and evaluating projects .... 30

Figure 5: The fundamental model for performance-based analysis of geometric design of

highways and streets (Ray et al., 2014) ................................................................................... 38

Figure 6: MnDOT family of transportation plans (MnDOT, 2017) ........................................ 47

Figure 7: Overview of MassDOT's project development steps (MassHighway, 2006) .......... 49

Figure 8: ODOT’s transportation system lifecycle (ODOT, 2017) ......................................... 50

Figure 9: From left to right (a) location page, (b) classification page, and (c) identification

page of MisplacedWheels.com (Pathak et al., 2021) ............................................................... 76

Figure 10: The user interface of Scooter Map (Lekach, n.d.) .................................................. 79

Figure 11: Zip codes with at least 20 scooters shown with a light blue shade. Thick border

depicts Portland’s boundaries. ................................................................................................. 82

Figure 12: Eight randomly sampled zip codes, shown in a darker shade of blue, that will be

used as the disjoint clusters for the study................................................................................. 82

Figure 13: The actual collected data overlaid on the zip code map after the data collection.

The map shows that all collected points are in accordance with the sampling plan as they are

within the dark blue shade. ...................................................................................................... 83

Figure 14: The number of violation and non-violation reports stacked on top of each other for

Portland and Seattle. The height of the bar is the total number of reports for each city. ........ 86

Figure 15: Count of micromobility vehicles audited in Seattle per scooter company operating

there, color coded by the company's colors ............................................................................. 88

Figure 16: Count of micromobility vehicles audited in Portland per scooter company

operating there, color coded by the company's colors ............................................................. 88

Figure 17: Number of vehicles audited, plotted next to the number of violations for each

micromobility company in Seattle ........................................................................................... 89

Figure 18: Number of vehicles audited, plotted next to the number of violations for each

micromobility company in Portland ........................................................................................ 89

Figure 19: Portion of different severities from 0 (meaning no violation) to 10 (severities 1

through 10 all mean some sort of violations) for Seattle ......................................................... 90

Figure 20: Portion of different severities from 0 to 10 for Portland ........................................ 90

Figure 21: Histogram and density plot showing the severity distribution of violating

micromobility vehicles in Seattle (severity 0 is not included) ................................................. 91

Figure 22: Histogram and density plot showing the severity distribution of violating

micromobility vehicles in Portland (severity 0 is not included) .............................................. 91

Figure 23: Distribution and severity of parking infraction reports in the Portland case study 94
Figure 24: Distribution and severity of parking infraction reports zoomed in on Downtown

and Central Portland areas for better visualization .................................................................. 95

Figure 25: Distribution and severity of parking infraction reports in the Seattle case study... 96

Figure 26: Micromobility parking infraction rates per census tract for the City of Seattle

shown with different colors (purple for highest rate bin and yellow for lowest rate bin).

Interactive version available at: https://rpubs.com/bornakhedri/micromobility-parking-

infraction-rate ........................................................................................................................... 97

Figure 27: Micromobility high severity parking infraction rates per census tract for the City

of Seattle shown with different colors (purple for highest rate bin and yellow for lowest rate

bin). High severity means only those with a severity greater than or equal to 7. Interactive

version available at: https://rpubs.com/bornakhedri/micromobility-high-severity-parking-

infraction-rate ........................................................................................................................... 98

Figure 28: Micromobility parking infraction rates per census tract for the City of Portland

shown with different colors (purple for highest rate bin and yellow for lowest rate bin).

Interactive version available at: https://rpubs.com/bornakhedri/micromobility-parking-

infraction-rate ........................................................................................................................... 99

Figure 29: Micromobility high severity parking infraction rates per census tract for the City

of Portland shown with different colors (purple for highest rate bin and yellow for lowest rate

bin). High severity means only those with a severity greater than or equal to 7. Interactive

version available at: https://rpubs.com/bornakhedri/micromobility-high-severity-parking-

infraction-rate ......................................................................................................................... 100

Figure 30: Zoomed in version for Figure 27, concentrating mainly on the Portland central

area. ........................................................................................................................................ 101

Figure 31: Zoomed in version for Figure 28, concentrating mainly on the Portland central

area. ........................................................................................................................................ 102

List of Tables
Table 1: Characterizing the literature by recommended goals for projects, dimension provided

for the performance measures, and list of performance measures provided............................ 31

Table 2: Results of informational interviews for identifying multimodal leaders based on U.S.

region and state/local jurisdiction ............................................................................................ 43

Table 3: List of documents for the selected four jurisdictions................................................. 44

Table 4: Alternative design processes and initiatives overlaps (Neuman et al., 2016) ........... 52

Table 5: Summary of the design processes for WSDOT, FDOT, MnDOT, ODOT, and

MassDOT, all compared to NCHRP 839's design processes guidelines ................................. 54

Table 6: Summary of design processes .................................................................................... 59

Table 7: Context classification comparison across jurisdictions ............................................. 60

Table 8: Design control comparison across jurisdictions ........................................................ 61

Table 9: Example of ODOT’s project-level performance measures by mode (ODOT, 2020b)

.................................................................................................................................................. 63

Table 10:List of MassDOT’s complete streets performance measures by mode (Lovas et al.,

2015) ........................................................................................................................................ 64

Table 11: Method(s) used by 13 U.S. cities for resolving micromobility parking infractions 72

Table 12: The results of interviews with five jurisdictions specifying key and good-to-have

features that they require a micromobility infraction reporter tool to incorporate .................. 72
Table 13: List of parking infractions for different cities grouped into 9 categories (SEA =

Seattle, PDX = Portland, SF = San Francisco, LA = Los Angeles, DC = Washington D.C.) . 75

Table 14: Linear regression models for Seattle census tracts ................................................ 105

Table 15: Linear regression models for Portland census tracts ............................................. 105

Table 16: List of micromobility infraction rules for Seattle, WA ......................................... 121

Table 17: List of micromobility infraction rules for Portland, OR ........................................ 122

Table 18: List of micromobility infraction rules for Redmond, WA ..................................... 123

Table 19: List of micromobility infraction rules for Spokane, WA ...................................... 123

Table 20: List of micromobility infraction rules for Boise, ID.............................................. 124

Table 21: List of micromobility infraction rules for San Francisco, CA ............................... 124

Table 22: List of micromobility infraction rules for Los Angeles, CA ................................. 125

Table 23: List of micromobility infraction rules for Washington D.C. ................................. 125
1 Introduction and thesis overview
We live in a world that is becoming increasingly multimodal where new mobility solutions

are also gaining popularity. It is important to address the needs of every mode in the planning

and design stages of a project, and to monitor the performance of various modes (including

new mobility modes) on project, corridor, or network levels. To gain better multimodal

outcomes from projects, engineers need clear guidance on what to do in the various phases of

project development and design. It is important to have a set of steps and processes that

engineers could look at and move forward with their design decision making. A review of

recent literature showed that, as this is a relatively new field, there needs to be more guidance

on different multimodal design processes and performance measures so that practitioners can

achieve better and more clear outcomes. In other words, there needs to be more guidance in

various phases of the project to answer questions such as: who the stakeholders are, how to

engage with the public, what design criteria to use, what performance measures to use to

evaluate projects, etc.? The first part of this thesis (Chapter 2) establishes a set of concrete

guidelines by reviewing the academic literature as well as practical guidebooks and

handbooks, and adopting the state of the knowledge and the best practices in this field.

Multimodal design processes, level of service methods, methods for evaluating projects

through performance measures, and a comparison of various jurisdictions’ design processes

are all reviewed. Chapter 2.2 shows the literature review and Chapter 2.3 illustrates the best

practices review.

Another area where this thesis seeks to contribute is to that of performance monitoring for

new mobility solutions, particularly micromobility. With micromobility vehicles (i.e., shared

e-bikes and e-scooters) arriving in cities, one rising issue is that mis-parked bikes or scooters

block sidewalks and pathways and interrupt or disrupt other modes such as pedestrians,

particularly those with disabilities. The second part of this thesis (Chapter 3) introduces a

1
performance monitoring solution for micromobility parking infractions. We talk about an

application that we developed which gathers crowdsourced data from city residents for

reporting micromobility parking violations. The motivation behind the app along with the

development is briefly described in this chapter. We also talk about different categories of

parking infractions by comparing the rules for eight different cities. We then report on a case

study done in the summer of 2020 where a data-driven parking audit was conducted in the

cities of Seattle and Portland. Some summary statistics along with spatial analysis of the data

are mentioned. We then discuss the use of publicly available data from the census and from

open city portals, to model the number and severity of infractions based on different aspects

of various neighborhoods including factors from the built environment (i.e., number of transit

facilities and number of bike racks) and socioeconomic and demographic factors (e.g.,

median age, median income, minority rate, etc.). The results of the model are mentioned, and

a conclusion is drawn upon in this section.

Lastly, in Chapter 4, we draw a general conclusion of both chapters to shape the

understanding of how transportation planning, design, and performance monitoring can go

hand in hand when preparing for a multimodal future in the era of emerging mobility

solutions.

2
2 A review of multimodal design guidelines, processes, and
performance measures
Acknowledgement: This chapter is based on a project funded by the Washington State

Department of Transportation in which they aimed to establish potential changes in

WSDOT’s design process that would more effectively support agency staff in making

decisions about the installation and design of multimodal improvements based on

performance. The first phase of this project included a national scale literature review of

academic papers and research reports to establish the “state of the knowledge”, and the

second phase included a review of federal and state guidebooks and design manuals to

establish the “state of the practice”. This chapter is a report on the first two phases of the

project.

2.1 Introduction

In this project, we aim to assess and improve the application of multimodal performance

measures and gain an understanding of multimodal design processes, to set forth a list of

recommendations for WSDOT (and other state DOTs) to incorporate into their design

manuals. As a first step, we are thoroughly reviewing available literature and documents in

this field. The review itself can be divided into two tasks:

1) To characterize the “state of the knowledge”: understand what scholarly research has been

done on achieving multimodal goals in projects. We searched for research on what processes

can support an effective multimodal design, and similarly, what the process is for effectively

developing performance measures to assess multimodal goals. We are likewise interested in

finding performance measures that have been suggested for measuring the multimodal

performance of a system.

3
2) To get a grasp of the “state of the practice”: to better understand what methods are being

used by different authorities (DOTs & local jurisdictions), what multimodal performance

measures they suggest, and processes that they have implemented in their project

development and design phases to support successful multimodal design in practice.

The second task will be achieved through the review of jurisdiction design manuals,

guidebooks, and handbooks; whereas the review for the first task is focused more on

scholarly literature and research. Chapter 2.2 provides a synthesis of the literature reviewed

for the first task and Chapter 2.3 synthesizes the review of the best practices for the second

task.

2.2 Literature review

In this section, we review literature related to multimodal level of service and methods for

evaluating projects through performance-based design and developing multimodal

performance indicators. This literature review builds upon the work of previous review

papers and reports including Dowling et al. (2008) and Lasley (2016) while also integrating

newer studies. A guide to the materials presented in this literature review is provided in

Figure 1.

First, we look at the traditional approach to transportation performance management, before

turning to the recent shift towards multimodal mobility. We review some key terminology,

and how the methods in the Highway Capacity Manual (HCM) have evolved over the years

and look at how Levels of Service are defined for different modes. We then turn to other

efforts at developing multimodal level of service (MMLOS) measures, and the development

of performance measures as an important part of the design process, including the difference

4
between supply- and demand-based performance measures. Finally, some conclusions are

synthesized based on the literature review.

The documents used for this literature review include journal articles, conference

proceedings, as well as research reports. Several other state-level handbooks and design

manuals were also identified during the search for literature; however, those will be discussed

in more detail in subchapter 2.3, that is the synthesis of the state-of-the-practice.

Figure 1: Literature review guide

5
2.2.1 Conventional approach to performance measurement

Traditionally, the Highway Capacity Manual (HCM) has been used for measuring the

performance and operations of highways and other transportation network facilities. Single-

mode measures such as delay and travel speed have been primarily used in the past, and are

still in use among many practitioners. Furthermore, these measures were primarily

established for measuring the performance of automobiles and were focused on automobile

movement; as a result, they have also resulted in projects that increase speed and expand

roadway capacity (Seskin et al., 2015). A recent study by Dock et al. (2017) found through

interviews with peer agencies that congestion and delay measures are persistently popular

among the public and policymakers.

2.2.2 A shift towards multimodal and reshaping communities

In recent years, a shift has begun in how some agencies are planning and designing their

facilities and how they are distancing themselves from the traditional approach of auto-

centric performance measurement. Federally, actions such as the enactment of the Intermodal

Surface Transportation Act (ISTEA) of 1991 initially shifted transportation policy from

building of the national highway network to integrating multimodal transportation systems

(Sinclair et al., 2019). Those policy transformations were continued by other national

legislation such as the Moving Ahead for Progress in the 21st Century (MAP-21) Act of 2012

(Sinclair et al., 2019). Parallel movements have also occurred outside of government, such as

the National Complete Streets Coalition. This nationwide movement launched in 2004,

aiming to integrate people and facilities in all dimensions of transportation networks

including planning, design, construction, operation, and maintenance. According to Seskin et

al. (2015), the National Complete Streets Coalition is “a non-profit, non-partisan alliance of

public interest organizations and transportation professionals committed to the development

6
and implementation of Complete Streets policies and practices.” More than 700 agencies

have adopted Complete Streets policies as of 2015.

The ramifications of auto-centric design policies and their adverse effects on the

transportation system and environment have led to the transitions described above.

Pedestrians and bicyclists have especially felt the unfavorable effects resulting from high

volumes of motorized vehicles going through urban areas faster than other modes (Sanders et

al., 2010). There has been a decrease in roadway safety, walkability, and bikability, while air

pollution caused by vehicle emissions has increased (Sanders et al., 2010).

Several studies have indicated that one of the toughest challenges for Departments of

Transportation (DOTs) and Metropolitan Planning Organizations (MPOs) in project

evaluation and decision-making is the lack of consistent and comparable multimodal

performance measures over time and across geographic regions (Wang et al., 2016). This

deprives professionals of a framework that measures progress towards the broad objective of

multimodality (Sanders et al., 2010). State DOTs routinely use performance measures to

assess transportation systems. However, since this assessment is mostly based on the

traditional highway engineering perspective of prioritizing automobiles, multimodal

measures have been applied in more limited contexts (Dock et al., 2017).

2.2.3 Definitions of performance measures, quality of service, and level of service

The words "quality of service", "level of service", and "performance measures" are often used

interchangeably; however, caution is needed since the three sets of terms each have different

meanings (Phillips et al., 2001). The first edition of the Transit Capacity and Quality of

Service Manual defines the terms as follows (Kittelson & Associates, Inc. et al., 1999):

Quality of Service: “The overall measure or perceived performance of service from

the passenger's or user's point of view.”

7
 Level of Service: “LOS is a range of six designated ranges of values for a particular

aspect of service, graded from “A” (best) to “F” (worst) based on a user's perception.”

Performance Measures: “A quantitative or qualitative factor used to evaluate a

particular aspect of service.”

Further, “service measures” is also another term used in this context, which differs from

performance measures in that it represents the passenger's or user's point of view specifically,

whereas performance measures can reflect any number of points of view. Service measures

should be relatively easy to measure and interpret in order to be beneficial to users. Lastly,

level of service grades (A-F) are typically developed and applied to service measures.

(Kittelson & Associates, Inc. et al., 1999; Phillips et al., 2001). Traditionally, however,

performance measures often have been viewed as the “operator’s” point of view, and they

have been more vehicle-oriented, including a variety of utilization and economic measures

(Phillips et al., 2001).

Level of service remains a vital communication tool for planners and engineers to

communicate the results of their technical analyses to decision-makers, elected officials, and

the general public in a way that is easy to understand and interpret (Dowling et al., 2002).

However, public agencies and operators should understand that it is important not to just

focus on calculating an LOS range because a host of other factors may influence the quality

of service, and those may not easily show themselves in an A to F classification (Kittelson &

Associates, Inc. et al., 1999; Phillips et al., 2001).

2.2.4 What is multimodal performance measurement?

According to Heller (2014), transportation performance measurement is defined in the

FHWA as a “strategic approach that uses system information to make investment and policy

decisions to achieve national performance goals”. The word "multimodal" is seen

8
increasingly in the literature; however, different people may use the term in different ways

and to represent different concepts (Sinclair et al., 2019). The performance of a multimodal

system can be expressed through measurement initiatives such as multimodal accessibility

(focusing on the ability to reach destinations), multimodal connectivity (quantifying the

cohesiveness of travel across modes), or multimodal productivity (the productivity of actual

person trips made across the network through all modes) (Sinclair et al., 2019).

Lasley (2016) argues that there are two main approaches to multimodal performance

measurement: (1) “a collection of single-mode measures bound in a single resource”; or (2)

“a single resource/measure that examines multiple modes simultaneously using a common

comparison factor”. Often times, multimodal performance measures are developed for

estimating progress towards a broader goal. However, there is no single broad goal for

multimodal performance measures since effective multimodal performance measurement

should include metrics that examine a variety of goals, e.g., system quantity, effectiveness,

efficiency (delay), connectivity, competitiveness, and safety (Lasley, 2016).

Lasley (2016) further examines recent relevant literature, providing a common level of

understanding and recommended steps to create useful multimodal performance measures.

The paper concludes that measures that examine network performance among and between

multiple modes are rare. Furthermore, Kanafani et al. (2010) and Cambridge Systematics

(2010) also mention that there is a need for measures that use a common denominator (e.g.,

delay, travel time, etc.) to allow the evaluation of the system and compare one mode against

the others to match the mode and route decisions.

Another description by Seskin et al. (2015) states that performance measures can generally be

interpreted to mean the data inputs that are needed when:

• Tackling long-term planning efforts;

9
 • Selecting projects to fund;

• Performing an alternatives analyses (evaluating all feasible options for a project);

• "Considering specific elements when finalizing a project’s design";

• Displaying the present state of a system using tools such as a dashboard; and Most
importantly,

• “Evaluating the outcomes of a built project”

According to Seskin et al. (2015), performance measurement is the process of “establishing

performance targets, modeling impacts, and monitoring results”. They point out that simple

before-and-after analyses may demonstrate how well a project achieved its intended goals for

elected leaders and residents. However, for transportation planners, designers, and engineers,

such measurement provides an additional benefit: “measuring the actual results of projects

allows them to make better-informed choices for future projects.

Furthermore, the scale of analysis matters when trying to apply the right performance

measures. Seskin et al. (2015) recommend that measures should be chosen thoughtfully and

with consideration of scale to avoid misinformation in decision-making and evaluating

results. For example, measuring vehicular LOS at only one intersection may lead the planner

or engineer to conclude that a wider intersection is needed, which may unintentionally cause

bottlenecks elsewhere on the corridor. Such inattention to scale will potentially reduce safety

and quality of the environment for individuals using other modes such as walking or

bicycling in a nearby intersection. On the contrary, if a particular segment is not yet

connected to a larger walking and bicycling network, perhaps doing a before-and-after

measure of the number of people walking or bicycling only on that street segment may be

misleading as the broader network level may be overlooked. Heller (2014) reinforces the

notion that scope and scale are important when considering performance measures, since

10
metrics can be applied at an entire transportation system level, a corridor level, or even a

facility level.

2.2.5 Review of multimodality in the Highway Capacity Manual (HCM)

2.2.5.1 HCM 1985

At the time of HCM 1985, the auto-centric design approach was at its peak, but there were

still some methods available for pedestrian and bicycle modes in the Highway Capacity

Manual. According to Phillips et al. (2001), HCM 1985 offered very limited scope for the

pedestrian and bicycle modes, and defined performance measures for the environments of

these two modes plainly as the degree of discomfort to the user due to overcrowding of the

facilities. The applicability of this sort of measure is extremely limited and is not suitable for

understanding the performance of those modes. Furthermore, multimodality was not

mentioned at that time in the HCM.

2.2.5.2 HCM 2000

HCM 2000, in addition to methods for automobile LOS, contained significant new methods

for analyzing pedestrian and bicycle LOS while also including a summary of the TCQSM for

transit analyses (Dowling et al., 2002). However, somewhat similar to the methods provided

in HCM 1985, these new methods use sidewalk and bicycle lane capacity for measuring

pedestrian and bicycle level of service and neglect the impacts of facility design and the

interaction of those modes with vehicular traffic on peoples' perceived level of service. Such

design features have a real influence on pedestrian and bicyclist satisfaction with a facility

and consequently its quality of service (Dowling et al., 2002).

For transit, however, due to the publication of the Transit Capacity and Quality of Service

Manual (TCQSM) (Kittelson & Associates, Inc. et al., 1999), several methods, performance

measures, and LOS measures were made available that took the perspective of transit riders

11
into account. These LOS measures were provided for transit systems, transit route segments,

as well as transit stops (Dowling et al., 2002).

2.2.5.3 HCM 2010

2010 was the year that multimodal design was highlighted in the manual. A study by

Elefteriadou et al. (2015) summarizes the multimodal aspects of HCM 2010. Their study

covers three different methods presented in the HCM for: (1) uninterrupted-flow facilities

(Chapter 15); (2) interrupted-flow facilities or urban streets (Chapters 16-18); and (3) off-

street pedestrian and bicycle facilities (Chapter 23). All are discussed below to understand the

HCM’s approach to multimodality.

(1) Uninterrupted flow

Uninterrupted flow refers to facilities where traffic is not controlled by traffic signals, stop

signs, or yield signs. A method for evaluating bicycle operations on multi-lane and two-lane

highways is provided in the manual. A bicycle LOS (BLOS) score scaled from A to F is

reported which represents "the quality of service for the bicycle mode when it is traveling

along the highway within the same right of way as other motorized vehicles." It is assumed

that bicyclists always use the rightmost lane of the highway (or shoulder when available).

The factors that the BLOS score is sensitive to are: “the vehicular demand in the rightmost

lane of the highway, the width of that lane, the width of the shoulder, percentage of heavy

vehicles, the speed limit of the facility, presence of parking, and pavement surface condition”

(Elefteriadou et al., 2015).

(2) Interrupted flow

12
HCM 2010 provides LOS calculation procedures for on-street transit, bicycles, and

pedestrians on urban streets. The manual uses multiple performance measures in determining

LOS, rather than the more-traditional single-measure approach (e.g., speed, delay, density).

Transit analysis for urban street facilities and segments

Transit LOS measures are provided in the HCM for evaluating on-street public transit service

in a multimodal context. On-street transit service comes in contrast to off-street service which

refers to transit in its own right of way (e.g., a transit lane) or transit that travels along a street

without stopping to serve passengers (e.g., express bus). The Transit Capacity and Quality of

Service Manual (TCQSM) provides more performance measures, computational methods,

and spreadsheet tools for evaluating capacity, speed, reliability, and quality of service of both

on- and off-street transit services. The method provided in the HCM simply provides a 6-

scale grade of A to F and is sensitive to the following: the frequency of on-street transit

service, perceived bus speed (includes variables like on-board crowding, reliability, and other

factors), and the quality of the pedestrian and waiting environments at bus stops.

Bicycle analysis for urban street facilities, segments, and intersections

The HCM provides BLOS for various scales and facilities, namely signalized intersections,

links (between intersections), segments (links plus intersections), facilities (multiple

consecutive segments), and off-street bicycle and shared-use paths/trails. The signalized

intersection BLOS - stratified as usual into LOS A to F - is sensitive to the following

variables: “lateral distance between the bicyclist and traffic, volume of traffic in the right

lane, percent of heavy vehicles, presence of on-street parking, and cross-street width"

(Elefteriadou et al., 2015). Bicycle speeds and volumes, and signal delay for the bicycles are

not factored in the bicycle LOS. No methods are provided for BLOS for unsignalized

intersections including roundabouts, due to lack of research.

13
BLOS is also calculated for links, which is sensitive to all the variables mentioned above

(except the width of cross-street since it is inspecting links and not intersections) in addition

to the following variables: the speed of traffic, the number of unsignalized access points, and

pavement conditions. Finally, for BLOS of urban street segments, a weighted combination of

the signalized intersection and link LOS values is used, and for the facility LOS, a weighted

average of the segment LOS values is utilized.

Pedestrian analysis for urban streets and intersections

Pedestrian LOS (PLOS) is also performed for the same geographic types/levels as for the

bicycle analysis. In addition to those levels, “street corner pedestrian storage and crosswalk

circulating capacity checks for pedestrians” are also provided in the HCM at a signalized

intersection. The PLOS methods provided in the HCM do not take into account ADA

accessibility; a separate assessment should be performed by the analyst to incorporate that

when evaluating the overall quality of service.

The PLOS at a signalized intersection is sensitive to: “pedestrian delay due to the signal,

lateral distance between the pedestrian and traffic, the volumes of traffic in the right lane, the

left and right turning volumes crossing the crosswalk while pedestrians have the walk

indication on, the percent of heavy vehicles, and the presence of on-street parking”

(Elefteriadou et al., 2015). PLOS also depends on the number of lanes on the cross-street as

well as the number of right turn channelization islands that the pedestrian has to cross.

Finally, the PLOS at the link level is sensitive to “the lateral and buffer separation between

pedestrians and traffic, the traffic volume in the right hand lane, the percent of heavy

vehicles, the speed of traffic, the presence of on-street parking and barriers such as street

trees, and the difficulty of making mid-block crossings (where legal)” (Elefteriadou et al.,

2015). Moreover, a density-based pedestrian LOS is calculated for sidewalks with high

14
pedestrian volumes, and the lower of the two LOS values is reported as the link LOS.

Segment LOS, similar to that of the bicycle analysis, gets calculated through a weighted

combination of the signalized intersection and non-density link LOS values; but, it is also

compared with the density-based link LOS and the worse is selected. Finally, facility LOS is

a weighted average of the segment LOS values.

(3) Off-street pedestrian and bicycle analysis

Methods are also provided in the HCM for analyzing off-street pedestrian and bicycle

facilities where three situations are considered: (1) a pedestrian only facility (e.g., stair,

pathway, etc.) where LOS is estimated by the available space for the average pedestrian; (2) a

pedestrian LOS on a shared-use path, where LOS is estimated by the number of times and

average pedestrian meets or is passed by bicyclists; and (3) bicycle LOS on a shared-use path,

determined by the number of times the average adult bicyclists meets, passes, or is passed by

other path-users.

As a conclusion, it is important to point that the HCM analysis provides performance

measures from a traffic operational quality viewpoint, and its provided methods estimate

performance measures particularly related to that aspect of transportation. Elefteriadou et al.

(2015) conclude that for comprehensive planning purposes, a wider set of performance

measures are needed that consider other aspects beyond that of traffic operational analyses,

such as environmental and financial measures.

2.2.5.4 HCM 6th Edition

Elefteriadou (2016) provides a brief introduction to the 6th edition of the Highway Capacity

Manual. This latest edition of the manual is subtitled “A Guide for Multimodal Mobility

Analysis,” pointing to the shift in perspective toward multimodal analysis. The manual

15
incorporates latest research findings into an extensive set of analysis tools for the operational

analysis of traffic.

Elefteriadou (2016) categorized four dimensions mentioned in the scope of the HCM6,

namely: capacity, quality of travel, quantity of travel, and accessibility. The HCM6 still

mainly focuses on evaluating the capacity and quality of service of facilities and modes

through data related to quantity, especially demand, as an input. They provide LOS as one of

the best-known measures for assessing the service of a facility; however, “they also provide

tools for estimating additional performance measures for a variety of modes and facilities.”

(Elefteriadou, 2016) Elefteriadou (2016) also clarifies that LOS measures are intended to

communicate about the operations of a facility to a non-technical audience and that

performance measures recommended in the HCM6 can be used by themselves without the

use of LOS.

Two new automobile-related chapters have been added to the manual, listed in Elefteriadou

(2016), including Chapter 11, Freeway Reliability Analysis and Chapter 17, Urban Street

Reliability and Active Traffic and Demand management. The focus of these new chapters

seem to be evaluating travel time reliability through the distribution of travel times, over a

broad period of time (e.g., a year) (Elefteriadou, 2016), and they do not seem to be involving

multimodal context into them. But, the study also mentions that “in response to the increasing

need to estimate the performance measures for pedestrian, bicycle, and transit facilities, as

well as the interactions with vehicles”, the HCM has provided certain methods for those

assessments namely, Chapters 16 through 23 (methods for assessing non-automobile modes

and their interactions with vehicles) and Chapter 24 (off-street pedestrian and transit

facilities). Chapter 15 of the manual evaluates bicycle operations on highways (multi-lane

and two-lane) with the addition of information from a newer research. The manual still

16
recommends to use TCQSM for the evaluation of transit facilities; however it does consider

the effects of transit along urban streets within a “multimodal analysis framework”

(Elefteriadou, 2016).

In summary, there have been some additions to the HCM6 as compared to HCM 2010; but

there does not seem to be a radical change in how HCM is calculating performance of

facilities.

2.2.6 Multimodal Level of Service: old and recent methods

In the late 1990s, some studies started addressing multimodality and the need for multimodal

performance measures in pursuit of the ISTEA legislation, and the Florida Department of

Transportation (FDOT) was one of the earliest funders of research in this area. Guttenplan et

al. (2001) was probably one of the first papers to study multimodal level of service analysis at

the planning level. The research focused on methods for determining the level of service

(LOS) to pedestrians, bicyclists, scheduled fixed-route bus users, and through vehicles on

arterials. It is based on work by FDOT, which had developed a multimodal LOS analysis

process to measure and provide mobility for diverse roadway users.

In another project, FDOT entered into a contract with the University of Florida and two

consulting firms that were the leaders of research in quality and level of service methodology

development at the time, to address the need for a planning level and quality of service

analysis for Florida (Phillips et al., 2001). Some of the objectives of the project (which were

shaped in the summer of 1998) were to: (1) perform a national literature review of

multimodal level of service methodologies to implement the best possible methodology in

Florida; (2) measure the performance of corridor segments in two districts by applying and

validating Bicycle LOS and Roadside Pedestrian Condition techniques; and (3) apply and test

new HCM performance measures for transit (in test districts).

17
The study further shows that there were some concerns across local governments, questioning

if statewide requirements will shift to standardizing and requiring use of multimodal

performance methods (Phillips et al., 2001). However, the focus at that time appeared to be

on developing techniques, not standardization, and to enable and support "local government

efforts for multimodal planning by offering professionally acceptable techniques." They

suggest several performance measures in the report and offer a literature review of processes

used up until that point (being now somewhat outdated, it is not included in this study).

In addition, Dowling et al. (2002) developed and tested a method that measured the user-

perceived quality of service from a multimodal perspective. The study also asserted that in

the HCM 2000 release, there was a shift towards analyzing automobiles and transit in the

same corridor for the first time; but, it indicated that there were not any efforts to compile the

auto and transit levels of service into an aggregate measure for corridor-level assessment.

The proposed methods in Dowling et al. (2002) estimate automobile and transit LOS analyses

based on the HCM 2000 and TCQSM, respectively, while calculating bicycle and pedestrian

levels of service based on models developed by the researchers. Four classes of corridors are

recommended, and the methods were tested on two classes of urban corridors, with and

without a freeway.

In more recent years, several multimodal level of service methods have become available

throughout the literature. One study used eight different MMLOS methods for an arterial

corridor section case study in Austin, Texas (Zuniga-Garcia et al., 2018). Zuniga-Garcia et al.

(2018) list the following methods:

1) Highway Capacity Manual;

2) Transit Capacity and Quality of Service Manual;

18
 3) Charlotte, NC, Urban Street Design Guidelines (USDG);

4) pedestrian and bicycle environmental quality indices (PEQI and BEQI);

5) assessment of level of traffic stress (LTS);

6) bicycle compatibility index (BCI);

7) deficiency index (DI); and

8) Walk Score, Bike Score, and Transit Score

The study is focused on evaluating pedestrian, bicycle, and transit modes in particular, and

comparing letter-metric grades (e.g., HCM and TCQSM) with the use of different score

measures (e.g., BCI, WalkScore, etc.).

Their major findings are that although the HCM MMLOS assessment is suitable, it requires

user training and significant data. Furthermore, TCQSM can be used hand in hand with HCM

to complement the transit mode for corridor evaluation. The authors also indicate certain

strengths and weaknesses of different approaches, for instance USDG being appropriate for

intersection analysis but not for corridor evaluation, and PEQI and BEQI being appropriate

for corridor analysis but not sufficient for robust intersection evaluation. They finally indicate

that they recommend applying the HCM and TCQSM methods to assess multimodal

performance in corridors. Also, they assert that there is still not an overall MMLOS that

includes multiple modes by any of the methods, and the way all these methods work is that

the analysis should be applied separately for each mode. Besides, an aggregation of the

results to provide an overall score requires knowledge and judgment of how to weight

different modes, making the aggregated mode results subjective and biased (Zuniga-Garcia et

al., 2018).

19
While several MMLOS methods are available in the literature, efforts for developing

heuristic methods have also been seen. Ni et al. (2013) proposed a new evaluation method

that captures both the efficiency and safety of all road users at the same time in one level of

service metric. MMLOS is used for evaluating delay (using methodologies similar to that of

Dowling et al. (2008)), and a multimodal safety level of service (MSLOS) is established for

evaluating safety through conflict checks at the intersection. They are both ranked 1 to 6. The

multimodal comprehensive level of service (MCLOS), is established, and a case study of a

typical intersection in the VISSIM-SSAM simulation platform assesses the method.

The work of Sinclair et al. (2019) focused on developing performance measures that help

assess the performance of a multimodal design project (e.g., an intersection, a street, etc.),

also known as facility-based performance measurement. The research discusses a Multimodal

System Productivity (MSP) score, which is the number of completed person trips per network

minute. This research is referring to "multimodal trips" in a network-level and trip-based

context (i.e., trips that use different modes along their path). This measure is defined based on

the definition of productivity: "the ratio of input to outputs in the production process". For

multimodal transportation systems, completed person trips are production output and network

travel times are production inputs; therefore, the MSP score is the number of completed

person trips per network minute.

Kanafani et al.(2010) identified the attributes that affect multimodal performance from the

demand and supply perspectives, and their roles in multimodal transportation. The research

offers a utility-theory-based analytical framework leading to the construction of an indirect

utility function that quantifies the user's perception of level of service for multimodal

alternatives. A set of metrics were also developed to quantify the measures of performance

from the supplier's standpoint. This work seems to emphasize the intermodal integration of

20
various modes (using an entire trip that consists of several links of different modes) rather

than focusing on what metrics can individually be used to measure the performance of each

mode for a certain project. Their utility function is based on the following factors for the

user's perspective: Time (access, waiting, in-vehicle traveling, transfer), money (out of

pocket, indirect, bundle), safety, reliability, and flexibility (Kanafani et al., 2010). The first

two are disaggregate factors (meaning that they vary for each mode-link of the entire trip)

whereas the latter three are aggregate factors (meaning that if even one of the mode links of

that entire trip is unsafe, unreliable, or inflexible, the user may disregard that entire

multimodal bundle as a whole). They also apply a utility function for the supplier's side,

which is broken down into the government's perception (issues of concern are equity, energy,

emission, monetary cost, and level of service) and the agency's perspective (factors are cost,

revenue, and subsidy).

2.2.7 Developing performance measures for project evaluation

There have been various studies in the pursuit of a true multimodal performance indicator,

but there have not been promising results. Sinclair et al. (2019) recently sought to conduct a

literature review to find an “ideal” multimodal system performance measure. They reported

conducting a detailed literature search and concluded that they did not find an “ideal”

multimodal transportation performance measure in their literature review, although the details

of their search were not mentioned in the report. They therefore developed their own measure

called Multimodal System Productivity (MSP).

Harvey et al. (2018) discuss the performance measure development process of a project by

arguing that many design guidelines provided by federal or high-level agencies differ in

terms of context with what is needed in a particular community. In other words, each agency

may have to develop context-specific design guidelines for their own use rather than directly

21
applying national or state-level guidelines. For instance, national or state design guidelines

may not be applicable or provide guidance on every design aspect of a complete street project

at a local level. Therefore, regardless of such design guidance being provided at various

higher levels, local agencies have to take action and develop their own “complete street

design guidelines”. If a region does not have complete street design guidelines for its locality,

Harvey et al. (2018) suggest that they determine their community needs and preferences

while examining existing street typologies, climate, and current and planned transport modes.

The study further anticipates that in the future, with more data collection, analyses, and

information documentation being readily available, complete street design guidelines will

likely change in consideration of what has and has not worked (Harvey et al., 2018).

There have been similar guidelines offered in the literature regarding project evaluation

through the development of objectives and performance measures. A study recommended

four general steps for agencies when undertaking project evaluation (Seskin et al., 2015):

1. Agree to goals and objectives of the project: the goals of the project need to be

established and agreed upon. This can be a challenging process since different needs are

expressed by residents, elected officials, and transportation leaders. However, challenges can

be overcome by encouraging participation and engagement in dialogue leading to consensus

and mutual understanding. Furthermore, certain goals may exist on a network-level rather

than a project-level. For example, a project by itself may not achieve many goals for a

particular street segment, but any contributions to a broader goal, when viewed in the context

of an entire funding program or the entire network, should be considered (e.g., there may be a

citywide goal to build a certain amount of accessible curb ramps each year).

2. Determine the best ways to measure goals: This step is to figure out which data

will bring about success. Asking community members about what they will want to know

22
about the project once completed is beneficial when setting up data-collection practices. It is

also recommended to explore alternative data sources as significant data collected by

agencies other than the transportation agency may be available (e.g., sales data from business

improvement districts, crash data from local police departments and hospitals, etc.)

3. Implement measures: Baseline data should be collected at this time for each

measure, and an appropriate timeframe for evaluation after completion should be established.

One best practice is to measure conditions a year before, a year after, and three years after the

project is completed. Continuous data collection may be needed for some measures. Notably,

not all measures need to be quantitative, and qualitative ones may be more relevant in some

cases (e.g., collecting quotes from people's experiences with the street and what they like

about it, either collected in the project outreach or through additional interviews after

completion).

4. Share results: Sharing results with the public is an even more powerful tool than

project evaluation by itself since it allows members of the public, elected officials, and other

partner government agencies to understand results. It is important to design and package an

attractive final product. Photos should be included, accompanied by quotes regarding the

project

Seskin et al. (2015) underscore that data can help make better decisions when developing

performance measures to suit the goals of a project, but it is not a substitute for community

vision. Furthermore, another recommendation is that performance measures do not have to be

complicated and sophisticated mathematical models; they can be as simple as “blocks of

sidewalk added” or “number of trees planted”.

Seskin et al. (2015) outline useful performance measures for seven common complete streets

goals, namely access, economy, environment, place, safety, equity, and public health. The

23
focus of these goals is on project-level evaluation, but some related network-level measures

are also included. The report further provides a very long list of performance metrics, which

we suggest that practitioners look at. The performance measures are separated by goal, mode,

and scale across seven tables in that document.

Lasley (2016) further agrees with previous research that agencies should consider asking

several questions that impact a multimodal performance measure before creating or deciding

on using that metric. These questions include:

1) “How will the measure be used?

2) What are the ultimate goals of measurement?

3) What exactly is being measured and does this match planning objectives?

4) Is each mode treated fairly within the measure?

5) Are both people and goods considered;

6) What are the likely and worst-case impacts of the use of the measure? and

7) Who is the audience and could they understand the measure?”

Furthermore, agencies should also examine which factors are already being measured well,

and which factors could use improvement (Lasley, 2016). Multimodal performance

measurement generally measures either one of six factors: 1- Quantity (volume/tonnage); 2-

effectiveness (throughput, such as person or vehicle throughput per hour or other similar

measures); 3- Efficiency (travel time or delay); 4- Competitiveness (reliability or cost); 5-

Connectivity (LOS measures or connectivity to other modes and/or facilities); and 6- Safety.

These share similar terms with those mentioned in (Elefteriadou et al., 2012).

24
The study further recommends that in developing performance measures, it is important to

make sure that highlighting one mode does not come at the cost of penalizing another

(Lasley, 2016). Moreover, agencies should consider performance measures that are easily

understandable and interpretable. The measures should not be too complicated to

communicate and perceive. Finally, agencies should set an expectation or expected outcome

for each mode. Similar to how roadways are classified in a different manner, expectations for

the travel speed we get from a freeway should be different than that of an urban arterial. We

should not expect the same standards across different modes (e.g., when comparing auto

delay to transit delay).

Elefteriadou et al. (2012) developed a framework to help agencies in selecting a set of useful

performance measures, consistent with their overall goals and quality of life desired by their

community. The implementation of this framework is in the following three steps:

1) Identifying goals and objectives for the project and the transportation system of the

region (e.g., reduce congestion, increase non-SOV travel, minimize environmental

impact, enhance safety);

2) select a set of measures appropriate for each goal from an extensive list of measures

(this list is available in Appendix A of Elefteriadou et al. (2012)); and

3) use evaluation criteria to assess each measure in light of the agency's goals, policies,

and resources.

Five mobility dimensions are identified: Measures of infrastructure and environment;

measure of demand and system utilization; measure of user perceptions; measure of safety;

and measure of sustainability (Elefteriadou et al., 2012), and are further categorized into

several subdimensions shown in Figure 2. For each of the subdimensions, an extensive list of

25
performance measures is identified in Appendix A of Elefteriadou et al. (2012). These

performance measures are not brought in this document as it spans 22 tables; we refer

practitioners to look at this important reference to identify their needed performance

measures.

Furthermore, there are seven broad mobility-related planning objectives included in their

study, which are then linked to the performance measures (for further information, refer to

Table B.2 of their report which provides performance measures by objectives and

characteristics). The planning objectives are: (1) Minimize ecological impact; (2) increase

accessibility; (3) increase non-SOV travel; (4) reduce congestion; (5) optimize freight

movement; (6) enhance safety; and (7) reduce air pollution. Finally, data requirements which

are needed for each indicator are assessed in the study, and are available in Appendix C of

Elefteriadou et al. (2012). Their study is probably one of the most useful pieces of literature

found in our review, and we therefore recommend that practitioners study it further and apply

their suggestions when implementing performance-based design (Elefteriadou et al., 2012).

Dock et al. (2017) state the importance of visualization tools for agencies to display how they

are addressing multimodal mobility. The study provides a set of measures and a data

visualization component for the District of Columbia DOT and for other agencies to use. The

performance measures include a set of commute-related metrics and are shown in Figure 3

based on their modal perspective and broad mobility category.

26
Figure 2: Classification of performance measures by mobility dimension (Elefteriadou et al.,
2012)

Figure 3: Performance measures by modal perspective and mobility category (Dock et al.,
2017)

27
The study highlights that “Washington State and Florida are leaders among states in

performance measurement” and that both of these states use congestion measures that address

multiple modes (Dock et al., 2017). The study specifically cites the Washington State DOT’s

Corridor Capacity Report, a web tool that evolved out of the Grey Notebook performance

publication (Washington State Department of Transportation, 2014). Congestion on major

Interstate corridors is described in The Corridor Capacity Report in terms of daily vehicle

delay; vehicle throughput; length and duration of routine congestion; and transit, park-and-

ride, and high-occupancy vehicle lane usage. Dock et al. (2017) conclude that this shift helps

the department consider its performance “through the lens of economic productivity”.

Similarly, the Florida DOT produces the Multimodal Mobility Performance Measures Source

Book, an annual report of mobility performance measures (Transportation Statistics Office,

Florida Department of Transportation, 2015) brought in Dock et al. (2017). A robust set of

mobility and system coverage measures are mentioned in the sourcebook that cover all modes

(i.e., all surface modes, aviation, and seaports). System performance is divided into four

broad categories: quantity, quality, accessibility, and utilization.

One study claims that transportation accessibility “is the measure that truly represents the

success of multimodal transportation systems from both the eyes of users and transportation

practitioners” (Tasic & Bozic, 2017). The paper further presents spatio-temporal accessibility

measures for pedestrians, bicyclists, and transit users using the City of Chicago as a case

study.

Sanders et al. (2010), similar to other named studies, recommends first identifying broad

objectives and goals for the project, and then adopting performance measures to suit those

goals. The study identifies five broad goals for a public agency (the study is particularly

making suggestions to CalTrans but they note that these can be adopted by other agencies as

28
well) including safety, mobility, delivery, stewardship, and service. They also identify

"complete and green street measures" which can measure progress towards the broad

objective of multimodality. They also set forth a list of measures which we highlight only a

number of them below.

For pedestrians and bicyclists, Sanders et al. (2010) proposed safety performance measures

including: pedestrian fatality rate per walking trip; pedestrian injury rate per walking trip;

bicyclist fatality rate per bicycling trip; and the number of pedestrian/bicyclist collision

hotspots on urban arterials. In addition, they suggest some key system mobility measures,

including: the ratio of sidewalk mileage to centerline roadway mileage, bidirectionally on

urban arterials; the ratio of Class II bicycle facility mileage to centerline roadway mileage,

bidirectionally on urban arterials; percentage of intersections that are ADA compliant;

number of pedestrian trips on urban arterials; and number of bicycle trips on urban arterials.

Figure 4 compares the recommended processes for project evaluation and performance

measure development, highlighting key themes mentioned by three of the most relevant and

thorough studies (Elefteriadou et al., 2012; Lasley, 2016; Seskin et al., 2015). Table 1

compares the recommended goals and objectives, along with dimensions for characterizing

performance measures, as found in a selection of the most relevant published papers.

29
Figure 4: Processes for developing performance-based measures and evaluating projects

30
Table 1: Characterizing the literature by recommended goals for projects, dimension provided
for the performance measures, and list of performance measures provided

31
2.2.8 Supply- vs. Demand-based performance measures
A recurring theme in the literature is the difference between supply- and demand-based

performance measures. Demand-based measures focus on how much a facility is serving

users (e.g., the number of bicycle trips on the urban arterial per day). In contrast, supply-

based measures emphasize the opportunities provided by facilities (e.g., ratio of bicycle

facility mileage to roadway mileage). Metrics such as Multimodal LOS are also performance

measures themselves although they tend to focus more on the supply-side of the facility and

“what it has to offer” rather than “how much it is being used”.

Lee & Miller (2017) emphasize the importance of differentiating between supply-based

measures and demand-based measures. Their study included a literature review on this topic.

They also argue that some methodologies want to measure what level of multimodality their

facility design is supplying to the user while others measure the level of usage of their facility

by different modes. For example, the “complete street score” is a supply-based measure that

evaluates how a facility serves pedestrian, transit, auto, and bicycle users, based on criteria

established by the community (e.g., a street passing through town should serve both transit

riders and bicyclists) (Kingsbury et al., 2011). A second example for a supply-based measure

is MMLOS provided in Dowling et al. (2008), which is more operational-focused and

evaluates auto driver, bus passenger, bicyclists, and pedestrians. Criteria such as pavement

conditions or lateral distance of bicyclist from drivers are considered in this methodology,

among other mainly supply-based characteristics of the corridor. On the other hand, an

example of a demand-based measure is indirectly implied by (Grant et al., 2012), where the

use of a facility by non-auto modes implies a greater level of multimodality.

Lee & Miller (2017) then imply that there is a lack of a suitable framework that can assess

multimodality through looking at supply and demand simultaneously. They use probability

32
theory and principal component analysis to create a new indicator based on both supply and

demand (i.e., modal shares and monetary investment for each mode).

Another study also notes the difference between supply- and demand-side methodologies and

implies that quality-of-service methodologies are considered "supply-side" assessments since

they are an evaluation of existing facilities (Phillips et al., 2001). A problem with supply-side

assessments is that they do not predict or estimate future demand. They are, however, very

valuable for decision-making purposes regarding investments. On the other hand, demand-

side methods are used to generate quantitative estimates of demand for multi-modal facilities,

such as methods that assess potential demand levels rather than actual demand. The study

further encourages the use of supply-side methodologies in coordination with some of the

demand-side assessments, especially when demand is associated with the quality of existing

facilities (Phillips et al., 2001).

Another topic includes looking at multimodal performance through the lens of achieving

intermodal integration (Kanafani et al., 2010). For achieving intermodal integration, there

needs to be a multimodal level of service measure or similar tool that shows the connections

between involved modes of a system (Kanafani et al., 2010).

For demand-based measures, a common theme mentioned in several studies was the lack of

data for measuring the true performance of a system in terms of facility usage. A study by

Barbeau et al. (2020) finds a solution in applying big data for improving transportation

measurement, particularly that of public transit through data sources such as GTFS and AVL.

Another study (Sinclair et al., 2019) also indicates that the use of smartphone data from

companies such as Google and Apple could help make up for the lack of facility-based data

when measuring multimodal network performance by using individual trip-based data from

33
smartphones rather than from facilities. However, there are privacy concerns when it comes

to disclosing those data, in addition to it not aligning with the companies' business models.

2.2.9 Conclusions

This literature review examined the history of multimodal approaches in transportation

design and methods for multimodal performance-based design. We found that in reaction to

decades of auto-centric design, there is an ongoing shift towards multimodal design. There

are a number of multimodal level of service methods and multimodal performance measures

available in the literature, however, none of them may serve as an “ideal” performance

measure since the needs of projects are inherently different from each other.

The approach suggested by several studies is that an agency should identify and define its

own performance measures for each project through evaluating their broad and project-

specific goals and the available data that they have. A summary of suggestions regarding

performance-based design are set forth below:

• The goals and objectives of the project should be identified as a first step.

These goals could be generic and broad terms such as safety, reliability,

utilization, etc. or could also be more specific such as "reducing pedestrian

fatality rate to amount X by year Z".

• Performance measures can be adopted through a readily available list for the

analysts to choose from. Two extensive lists of performance measures are

available in the literature, throughout Seskin et al. (2015) and in Appendices A

& C of Elefteriadou et al. (2012).

• Once performance measures are selected, the analysts should identify required

data sources to reach those goals.

34
• One mode should not be sacrificed against the other.

• Performance measures used for a project should be a combination of supply-

based and demand-based measures, not just focusing on one side in favor of

the other

35
2.3 Review of best practices

The second task of the project involved getting an understanding of the state of the practice in

multimodal and performance-based design. We first review certain definitions and

terminology from the viewpoint of more practical documents. We then highlight the results of

our interviews with several officials from DOTs and industry experts to see which states are

multimodal leaders and what documents are especially relevant. Then, we report on some of

those documents and synthesize the best practices outlined in them for finding performance

measures, conducting trade-off examinations, data required for performance measures, low-

cost data acquisitions methods, and general design processes.

2.3.1 Key terminology and antecedents to multimodal design

2.3.1.1 Practical design

The concept of practical design started in 2005, when the Missouri DOT stated their new

strategic objective, to “build good projects everywhere – rather than perfect projects

somewhere” (Neuman et al., 2016). Financial pressure led to the innovative approaches in

project delivery and funding (Neuman et al., 2016). The term practical design is defined as

"an approach to road and street engineering that prioritizes economy and seeks to optimize

return on capital investment across the entire highway program and system. Its goal is a use

of public funds that results in the best possible highway system" (MnDOT, 2018). This is not

necessarily a new or unique concept as there has always been careful attention to economy

and investment in engineering. The practical design movement is a "rediscovering of and

refocusing on these principles", and is becoming more important due to downward trends in

revenue and investment (MnDOT, 2018). Some natural outcomes of this movement are that it

encourages design flexibility and considering a variety of design values, using evidence-

36
based techniques that will yield high returns on investment, and financial sustainability

within the long-term fiscal outlook (MnDOT, 2018).

2.3.1.2 Performance-based design

Another increasingly popular term in recent years is performance-based design. Performance

measures were defined in the previous literature review section; nevertheless, it is important

to revisit these definitions from the standpoint of practical documents. Minnesota Department

of Transportation (2018) defines performance-based design as "a design that involves

determining design features in order to achieve desired outcomes and solve identified

problems, based on known direct effects of physical roadway features on actual

performance." Cambridge Systematics, Inc. (1999) also defines performance measurement as

"the use of statistical evidence to determine progress toward specific defined organizational

objectives". NCHRP Reports 785 (Performance-Based Analysis of Geometric Design of

Highways and Streets) and 839 (A Performance-Based Highway Geometric Design Process)

are also important practical documents that guide practitioners about performance-based

design processes.

Combining the two aforementioned concepts results in Performance-Based Practical Design

(PBPD). This approach is the use of performance-based processes and methods to solve

problems while realizing there are limited monetary resources and that public funds must be

spent wisely with a long-term, system-wide outlook (MnDOT, 2018).

NCHRP Report 785 sets forth a fundamental model for understanding the desired outcomes

of a project, establishing and selecting performance measures that align with those outcomes,

evaluating the performance of various alternative geometric design decisions, and reaching

solutions that achieve the desired outcome while ensuring they are financially feasible (Ray

et al., 2014). This process is shown in Figure 5. NCHRP Report 839 points out that

37
throughout history, design decisions have largely been based on specified design values and

dimensions, and that highway and street design processes rely on standards that set minimum

design feature values or ranges of values (Neuman et al., 2016). These dimensions that came

from physical or mathematical models at the time, were intended to provided operational

safety, efficiency, and comfort for the road user and were once thought to automatically

provide good performance, while it is difficult or impossible for a designer to quantitatively

display how the facility will perform based on these dimensions (Neuman et al., 2016). For

example, minimum values or ranges for a facility’s lane width that come from a table based

on other dimensions such as design speed, do not guarantee good performance and do not by

themselves quantitatively characterize how the project will perform and satisfy goals such as

throughput, travel time, etc. In other words, engineers need performance metrics to measure

their progress towards their project goals and objectives

Figure 5: The fundamental model for performance-based analysis of geometric design of

highways and streets (Ray et al., 2014)

A standard dimension that might have once been suggested as a performance goal, can no

longer necessarily be considered as a desired outcome (Ray et al., 2014). Performance-based

38
practical design focuses on actual functional performance. Documents such as MnDOT’s

PBPD guidelines define broad performance characteristics such as: quality of service, safety,

reliability, accessibility, infrastructure integrity, ease of use, ease of maintenance, visual

quality, fit to context and community (MnDOT, 2018).

2.3.2 Informational interviews for identifying leaders in multimodal design

We identified multimodal planning and design experts from a list of participants in the

Transportation Research Board 2021 Annual Conference's multimodal subcommittee virtual

meeting. We sent 13 emails to a combination of USDOT (FHWA), state DOT, local DOT,

and industry experts, asking for 15 minutes of time to talk about agencies and jurisdictions

that are considered multimodal leaders and recommendations of state or local design manuals

that are especially relevant to our research. We received nine responses that consisted of five

state DOT officials and four industry experts, whose names and employers are kept

anonymous.

The interviewees helped identify local or state jurisdictions that had done previous work

regarding multimodal design, performance-based design, context-sensitive solutions, and

complete streets, and/or had some level of documentation for these topics in their design

manuals, handbooks, or guidebooks. All but one state DOT interviewees, in addition to

identifying other state or local multimodal leaders, identified their own agency as having

done some level of work in these fields. On average, each DOT representative identified

about seven jurisdictions as multimodal leaders (including themselves), whereas each

industry expert identified about 8 leaders on average (particularly, one of the experts

highlighted 19 jurisdictions leading to a higher average).

Our interview protocol began with a brief two-minute introduction of the research project and

then moved to five open-ended questions, although not every interview included all five

39
questions. The discussion for some questions was more detailed and automatically led to the

responses we were looking for to our other questions. We specified that by multimodal, we

meant active transportation (e.g., pedestrians, bicycles), freight (e.g., trucks, delivery vans,

etc.), transit (e.g., buses, trains, paratransit), and single-occupant vehicles, and asked them to

incorporate information about all these modes in their response. The questions then asked

were as follows:

• Are there any ongoing activities for expanding and pushing multimodal

transportation forward in your own state? If so, could you elaborate on those

efforts?

• Where could we learn more about these efforts in your state (i.e., online resources,

particular design manual chapters, etc.)?

• Are there other design manuals or documents belonging to other states that could

be identified as especially relevant to our research?

• In particular to performance measures, and data needs for different modes in

multimodal transportation, what resources can be identified?

• Do you have examples in your community where this new multimodal approach

was implemented? Has it made a difference and was it a helpful difference or not?

Table 2 shows the results of our interviews. The table shows the four jurisdictions that were

ultimately chosen as multimodal practice leaders to compare to WSDOT, with their votes

highlighted. Massachusetts had the highest number of votes with seven, Minnesota was next

with six, Florida had five, and Michigan had four. However, we elected not to examine

Michigan because Minnesota is in the same region and had more votes. Instead, we selected

as the last state Oregon, which had three votes. There were also three other jurisdictions with

40
three votes, i.e., Wisconsin DOT, Washington DOT, and Portland, OR. Wisconsin was also

not chosen because there was already Minnesota, chosen from the Midwest. Washington was

not chosen as this project was conducted for WSDOT and we aimed to identify other

jurisdictions. Portland, OR, however, was a viable choice but Oregon DOT was determined a

better fit given its status as neighboring state transportation department.

Interview highlights include:

• At least two of interviewees pointed out to Massachusetts DOT's (MassDOT)

Bike Facilities Guide and that it stood out among other practical guidelines.

One interviewee noted that Massachusetts adopted context-based design

earlier than other jurisdictions, continues to improve their Project

Development and Design Guide, and excels at the context-sensitive and

multimodal aspects of design.

• Florida DOT (FDOT) was acknowledged as a leader for their efforts to

transition to context-based design and context-sensitive guidelines. FDOT has

a thorough context classification document and has held several workshops in

recent years including workshops for complete streets and context-based

design.

• Minnesota was praised for its bike facility guidelines, its performance-based

practical design guidelines, as well as its documentation efforts, and Oregon

DOT was especially praised for their Blueprint for Urban Design (BUD).

• Interviewees identified two NCHRP reports as especially relevant. NCHRP

Report 785 (Ray et al., 2014) provides a principles-focused approach that

looks at the outcomes of design decisions as the primary measure of design

41
 effectiveness. The report presents methods to incorporate performance-based

analysis into the project development process. NCHRP Report 839

recommends a new highway geometric design process that is more aligned

with the current expectations of transportation agencies and communities by

focusing on transportation performance rather than the selection of values

from tables of dimensions.

• One interviewee noted that a particular county they know of is moving entirely

away from functional classification of roadways towards context classification

exclusively.

• Finally, one expert noted the decline in the incidence of performance measures

in design manuals and greater emphasis on design processes and tailoring

measures to projects.

42
Table 2: Results of informational interviews for identifying multimodal leaders based on U.S.
region and state/local jurisdiction

43
2.3.3 A review of jurisdictions’ available documents in multimodal, performance-
based, and/or practical design

Table 3 shows the key documents referenced in our interviews for each of the final four

states. We next summarize what distinguishes these documents as “best practice”.

Table 3: List of documents for the selected four jurisdictions

Document
FDOT MnDOT MassDOT ODOT
type

- FDOT Design Manual

(FDOT, 2020a) - Project Development - Highway
Main Design - Road Design Manual
and Design Guide Design Manual
Manual (MnDOT, n.d.)
- Florida Greenbook (MassHighway, 2006) (ODOT, 2012)
(FDOT, 2016)

- Context Classification
- Performance-Based - Separated Bike Lane
Guidelines (FDOT, 2020b)
Practical Design Planning & Design
- Blueprint for
Process and Design Guide (MassDOT,
- Transit Facilities Urban Design
Guidance (MnDOT, 2015)
Other Design Guideline (FDOT, 2017) (BUD)
2018)
Guidelines Volumes 1 & 2
- Guidelines for the
- Context Classification (ODOT, 2020a,
- Bicycle Facility Planning and Design
Framework for Bus 2020b)
Design Manual of Roundabouts
Transit (Kittelson &
(MnDOT, 2020) (MassDOT, 2020)
Associates, Inc., 2020)

- Complete Streets
Implementation Plan. - Statewide Bicycle,
- Oregon
Transportation M2D2: Multimodal Pedestrian,
Transportation
Plans Development and Delivery Multimodal, and other
Plan
(FDOT & Smart Growth Transportation Plans
America, 2015)

2.3.3.1 FDOT

The FDOT Design Manual (FDM) serves as the main design guide for the State of Florida.

Chapter 1 of the design manual covers project development and Chapter 2 covers design

criteria. The FDM replaced FDOT’s Plans Preparation Manual (PPM) (circulated since 1998)

in January 2018. The department has another important document titled FDOT Context

Classification Guidelines (FDOT, 2020b) that classifies Florida's roads into eight different

contexts. They also have policies, guidelines, and implementation plans for Complete Streets

44
projects. The FDM also has subchapters on pedestrian facilities (222), bike facilities (223),

shared-use paths (224), and public transit facilities (225). Intersections are in chapter (212)

and modern roundabouts in chapter (213).

Florida DOT leads in its use of context classification and context-sensitive solutions (CSS) as

well as their complete streets program. Interviewees noted that Florida moved to context

classification earlier than several other states and that their detailed CSS guidelines provide

state-of-the-art guidance for their engineers and practitioners in designing facilities that serve

all modes according to the land context. The Florida Greenbook (FDOT, 2016) provides

uniform minimum standards and criteria for the design, construction, and maintenance of all

public streets and highways, including pedestrian and bicycle facilities, in the State of

Florida.

FDOT was also able to fit its bike and transit guidelines within the CSS framework according

to one of our interviewees. Their public transit office has several documents that provide

technical guidelines in transit facilities designed to facilitate transit operations on and off the

roadway system. Their Context Classification Framework for Bus Transit (Kittelson &

Associates, Inc., 2020) provides illustrations pointing out the basic and desired elements for

transit facilities (i.e., bus lanes, bus stops, etc.) with respect to Florida’s eight different

contexts. Furthermore, their Transit Facilities Guidelines (FDOT, 2017) provides design

drawings for various transit facilities. FDOT updates their design manual rapidly, bringing in

newer guidance for all transportation modes, which is another reason they are identified as

multimodal leaders.

45
2.3.3.2 MnDOT

MnDOT has a Road Design Manual (RDM) that is now being transitioned into its successor

document, the Facility Design Guide (FDG). Both publications are now active on their

website, and as new parts of the FDG are published, the corresponding RDM parts will be

removed from the website, with linked FDG reference substituting their place. However, the

FDG is not complete yet, and therefore, the RDM was used to characterize their design

processes. They also have a Performance-Based Practical Design Process and Design

Guidance (MnDOT, 2018) document that provides information on what performance-based

design and practical design are and what they intend to do. This document addresses why

these approaches are needed today and offers some details of how to implement this

approach. However, it does not provide a full set of performance measures for project

evaluation.

MnDOT also has a Bicycle Facility Design Manual (MnDOT, 2020) which complements the

RDM with guidance on planning, designing, and maintaining bicycle facilities. The state also

has dozens of two to three page “infosheets” on such topics such as a paved shoulders and

side paths, each providing concise guidance on a particular matter. Interviewees

complimented MnDOT’s effective documentation and practice of publishing this guidance on

their website.

The department also has several statewide system plans, including a Statewide Bicycle

System Plan, Pedestrian System Plan, Transit Plan, Freight Plan, ADA Transition Plan, and

more importantly a Statewide Multimodal Transportation Plan. An overview of the DOT's

plans is shown in Figure 6. All these plans provide some level of insight into the future and

establish the visions and goals that the state has for the years ahead. In particular, the

Statewide Multimodal Transportation Plan (MnDOT, 2017) is a 20-year plan based on the

46
Minnesota GO vision and is the highest level policy plan for all transportation modes that

aims to maximize the health of the people, the environment, and the economy. The document

outlines the Minnesota GO Vision, discusses the current state of the state's transportation

system, and reviews key trends in the state's economy, population, environment, and

transportation plan. The document also provides guidance on how they plan to move forward

by presenting objectives, performance measures, and strategies to move towards their two-

decade vision. The performance measures provided in the document are mostly broad, state-

level measures such as system reliability and delay measures for the interstate and national

highway system, average annual aircraft delay, annual transit on-time performance,

percentage of state-owned sidewalk miles compliant with ADA standards, annual greenhouse

gas emissions from the transportations sector, etc. All of these measures look at the

transportation system as a whole, rather than serving as performance measures for one single

project (MnDOT, 2017).

Figure 6: MnDOT family of transportation plans (MnDOT, 2017)

47
2.3.3.3 MassDOT

Massachusetts DOT was the jurisdiction with the most votes as a multimodal practice leader

in our interviews. Several interviewees pointed to their innovative project development and

design guidelines which take both multimodal and context-based design into consideration.

MassHighway’s Project Development and Design Guide (MassHighway, 2006) sets several

goals for the project development process from concept to construction, including: (1) to

ensure context sensitivity through open dialog between project proponents, reviewers, the

public, and other parties; (2) to think beyond the pavement to achieve the optimum

accommodation for all modes; (3) to encourage early planning, public outreach, and

evaluation to identify project needs, objectives, issues, and impacts before expanding

significant resources; (4) to “achieve consistent expectations and understanding between

project proponents and those entities who evaluate, prioritize, and fund projects”; and (5) to

ensure resources are allocated to projects that address local, regional, and statewide priorities

and needs.

MassDOT’s Separated Bike Lane Planning & Design Guide (MassDOT, 2015) was also

mentioned by several interviewees who noted the guide was ahead of its time when released

in 2015. This document supplements the bicycle facility design advice in the Project

Development & Design Guide (Chapters 5 and 6) by providing guidance on where to

implement separated bike lanes as well as how to design them as part of a safe and

comfortable network of bicycle facilities. In addition, MassDOT’s Guidelines for the

Planning and Design of Roundabouts was referred to several times by our interviewees.

MassDOT's design manual, titled "Project Development and Design Guide", essentially

divides a project into two parts: project development and design. Chapter 1 of the book has

some general introductions, and Chapter 2 talks about the project development step, whereas

48
Chapters 3 to 14 are basic design chapters. Chapter 3 talks about basic design controls, while

Chapter 4 establishes parameters for designing horizontal and vertical alignments of streets

and highways. Chapter 11 has info on shared-use paths and greenways. Chapter 12 has info

on intermodal facilities and rest areas including park & ride facilities and transit centers.

Other chapters include intersection design, interchanges, bridges, access management, etc.

They develop an eight-step project development process to move a project from problem

identification to completion, which can be seen in Figure 7.

Figure 7: Overview of MassDOT's project development steps (MassHighway, 2006)

49
2.3.3.4 ODOT

ODOT uses two primary sources for their design: 1) the ODOT Highway Design Manual;

and 2) the Blueprint for Urban Design (Volumes 1 & 2). These two documents complement

each other and are used by all ODOT engineers for designing facilities. Our interviewees

appreciated the Blueprint for Urban Design for having innovative guidance on urban contexts

and roadway classification, design flexibility, and a multimodal decision-making framework.

ODOT also uses its Project Delivery Guide (ODOT, 2017) to provide step-by-step guidance

in the development and delivery of projects from transportation planning to constructing

management transition. Figure 8 displays the ODOT Transportation System Lifecycle.

ODOT also relies on the Oregon Transportation Plan (OTP) to provide a system-wide context

for project selection and design. There is an OTP amendment on performance measures that

has guidance on performance-based statewide transportation planning processes. In addition,

Oregon Highway Plan includes federal requirements for tracking certain performance

measures.

Figure 8: ODOT’s transportation system lifecycle (ODOT, 2017)

50
2.3.3.5 NCRHP Report 839

NCRHP Report 839: A Performance-Based Highway Geometric Design Process provides a

geometric design process for highways that focuses on the transportation performance of

facilities rather than design dimensions. The report starts by reviewing a history of highway

design from the 1940s up until the 2010s and shifts in project needs and design processes for

10-year periods. The report describes a movement towards more flexibility in design to help

transportation projects meet the needs of multiple stakeholders, and due to these shifts, the

emergence of alternative design concepts that have become part of the practice (Neuman et

al., 2016). These emerging concepts include the: (1) complete streets concept; (2) concept of

Context Sensitive Design - CSD (now often referred to as Context Sensitive Solutions -

CSS); (3) concept of performance-based design; (4) concept of practical design; (5) the

design matrix approach; (6) the safe systems approach; (7) concept of travel time reliability;

(8) concept of Value Engineering (VE); (9) concept of designing for 3R (Reconstruction,

Rehabilitation, Resurfacing) projects; and (10) concept of design for Very Low-Volume

Local Roads (VLVLR) (roads with ADT ≤ 400) (Neuman et al., 2016). Each of these

alternative design concepts and how they achieve their goals are briefly described in the

report. Table 4, taken from Neuman et al. (2016), lists important insights from these design

processes and illustrates how they compare and overlap across these various initiatives. One

key takeaway from this table is that from the 10 alternative design processes, only two

highlight that roads serve all road users and not just motor vehicles (multimodal design),

namely, complete streets and CSS. The report describes a new geometric design approach

informed by these various concepts that focuses on transportation performance.

51
 Table 4: Alternative design processes and initiatives overlaps (Neuman et al., 2016)

2.3.4 Comparison of project development and design processes through key design
steps and elements

In this section we compare the design processes for the four peer states, WSDOT’s process,

and the design process identified in NCHRP Report 839 together. Table 5 gives a summary of

the project development and design processes with the left-most column showing the 11-step

process from NCHRP 839 and the remaining columns showing how the states map to this

framework. The table shows that the design processes for all jurisdictions share some

common traits while some have more details than others. Table 6 shows a summarized

version of which jurisdictions follow which steps, including two more steps for a total of 13.

52
Table 7 shows a comparison of the various context classifications that these states describe.

Lastly, Table 8 compares the design controls of the various jurisdictions.

All the jurisdictions start their design by identifying a need or problem. They work to clearly

develop a need statement in the first step without prescribing a solution at the start. In the

next step, the design lead identifies and charters a group of project stakeholders including

those internal and external to the agency. This step also involves public outreach for some of

the jurisdictions.

The project team refines and confirms the problem or need statement to then inform the

development of alternatives and the selection of the preferred project scope. The scope will

establish the project type such as a new road, reconstruction, or 3R (reconstruction,

rehabilitation, and resurfacing). Next, the team evaluates the project context and geometric

design criteria using a context-based classification or context-sensitive. The team identifies

design controls such as target speed, traffic volumes, LOS, and road user attributes. Next, the

team applies the appropriate geometric design process and criteria to the project.

Once the geometric design is complete, an inclusive and interdisciplinary team evaluates

geometric alternatives that address the need or solve the problem, within the context

conditions and constraints. The team then makes key design decisions and generates

documentation to inform the final design decisions before transitioning to final engineering.

In the aftermath of the project, continuous monitoring and feedback allow the responsible

agencies to evaluate the project’s performance relative to the design goals.

53
 Table 5: Summary of the design processes for WSDOT, FDOT, MnDOT, ODOT, and MassDOT, all compared to NCHRP 839's design processes guidelines
No. NCHRP 839 WSDOT FDOT MnDOT ODOT Mass DOT
1 Define the • Clearly identify the Fully define and document The earliest step in the design Identify the problem and need for the Step I: Problem/Need/Opportunity
Transportation baseline need. Define it in the objectives of the project process is determining project project Identification
Problem or Need terms of performance, and the scope of activities purpose, need and problems The proponent completes a Project Need
contributing factors, and to accomplish them (FDM followed by establishing Form (PNF). This form is then reviewed
underlying reasons for the 110.2) desired outcomes and goals. by the MassHighway District office which
baseline need (Chapter (MnDOT PBPD) provides guidance to the proponent on the
1101 of design manual). subsequent steps of the process"
• Identify the objective, in Some key principles that (MassHighway PDDG Exhibit 2-11)
simple, direct terms. MnDOT highlights for
• Identify the successful project
performance metric(s) development are:
involved. • Balance safety, mobility,
• Include one or more community, and
quantifiable statements. environmental goals in all
• Exclude any description projects.
or discussion of potential • Involve the public and
solutions. affected agencies
• Address all modes of travel
• Apply flexibility inherent in
design standards (MnDOT
RDM p. 2-1(1))
2 Identify and Charter Engage local partners and Public Information and Understand the classification
All Project stakeholders at the Outreach (FDM 104.2) and of various highways
Stakeholders earliest stages of scope Community Awareness according to who has
• Internal Agency definition to account for Plan (FDM 104.3): responsibility for its
Stakeholders their input at the right • Identify partners maintenance, improvement,
• External Agency stage of the development • Identify project and traffic regulation
Stakeholders process. stakeholders (MnDOT RDM p. 2-2(1) ):
• Other External Engage with the • Identify target audiences • Jurisdictional systems:
Stakeholder Groups community to fully • Develop the message(s) Trunk highway system,
or Agencies understand: • Determine County Highway system,
• Directly Affected • Performance gaps communication strategies township road system,
Stakeholders • Context identity (more in the Florida Public Municipal city street system
• Stakeholder • Local environmental Involvement Handbook) • State aid systems: County
Chartering issues • Determine State Aid Highways (CSAH)
• Modal priorities and communication timing and Municipal State Aid
needs (Chapter 1100) Streets (MSAS).

54
No. NCHRP 839 WSDOT FDOT MnDOT ODOT Mass DOT
3 Develop the Project Translate the need into The Department’s project • Project scoping: Identify Step II: Planning
Scope: Refinement specific performance manager is responsible for system deficiencies and needs "Project planning can range from
and Confirmation of metrics and select targets the development, review through operation monitoring, agreement that the problem should be
Problem or Needs in accordance to what the and approval of the project data from management addressed through a clear solution to a
Statement design shall achieve. A objectives, scope of work, systems (bridge, pavement, detailed analysis of alternatives and their
contributing factors and schedule in accordance safety, etc.), maintenance impacts." (MassHighway PDDG Exhibit
analysis (in Chapter with the Project problems, and public 2-11)
1101) refines focus in Management Guide. They comments. (MnDOT RDM p.
order to resolve the must also verify that 2-4 (1)). Step III: Project Initiation
specific performance required funds are in the • Consider (1) Project "The proponent prepares and submits a
problems and helps work program. (FDM) programming; (2) Cost- Project Initiation Form (PIF) and a
define the potential scope effectiveness policy; and (3) Transportation Evaluation Criteria (TEC)
of project alternatives. Value engineering form in this step. The PIF and TEC are
informally reviewed by the Metropolitan
Planning Organization (MPO) and
MassHighway District office, and
formally reviewed by the PRC."
(MassHighway PDDG Exhibit 2-11)

4 Determine the Three basic types of • The objectives and available Types of Projects:
Project Type and projects: funding for the project must • Modernization [New
Design • New Construction be balanced early in the Construction/Reconstruction (4R)]
Development • Add Lanes and scoping process. There are • Preservation [Interstate
Parameters. Project Reconstruction three investment categories: Maintenance/Resurfacing, Restoration,
types are: • Other Projects (RRR, (1) New construction / and Rehabilitation (3R)]
• New Roads operational improvements, reconstruction, (2) • Bridge
• Reconstruction safety enhancements, etc.) Preservations, and (3) • Safety
Projects Preventive maintenance • Operations
• 3R (MnDOT RDM p. 2-5 (2)) • Maintenance
(Reconstruction, • Miscellaneous/Special Programs
Rehabilitation, and • Single Function
Resurfacing) • ODOT Resurfacing 1R (ODOT
Projects HDM 1-15)

55
No. NCHRP 839 WSDOT FDOT MnDOT ODOT Mass DOT
5 Establish the Identify the land use and Establish and document the • Functional classification: (1) Roadway Classification: (1) Statewide • Roadway Context (MassHighway
Project’s Context transportation context project's surrounding Arterials, (2) Collectors, and Highways, (2) Regional Highways, (3) PDDG Section 3.2): Area types: Rural,
and Geometric (which includes context: (FDM) (3) Local roads and streets District Highways, (4) Local Interest Suburban, Urban & Roadway types:
Design Framework environmental use and • Functional Class: (1) (MnDOT RDM p. 2-5(1)) Roads (ODOT BUD) Freeways, major arterials, minor arterials,
• Develop Project constraints) for the Limited Access Facilities, • Types of highways: (1) major collectors, minor collectors, local
Evaluation Criteria location (Chapter 1102). (2) Principal Arterial, (3) Two-lane, (2) multi-lane Context classifications: (1) Traditional roads and streets
Within the Context • Land Use Categories: Minor Arterial, (4) highways, and (3) scenic Downtown/ Central Business District
Framework (1) Rural, (2) Suburban, Collector, (5) Local Roads byways (MnDOT RDM p. 2- (CBD), (2) Urban Mix, (3)
• Establish (3) Urban, and (4) Urban • Context Class: C1: 5(3)) Commercial Corridor, (4) Residential
Decision-making Core Natural, C2: Rural, C2T: Corridor, (5) Suburban Fringe, (6)
Roles and • Transportation Context: Rural Town, C3R: Rural Community (ODOT BUD)
Responsibilities (1) Roadway type, (2) Suburban Residential,
• Determine Basic Bicycle route type, (3) C3C: Suburban
Geometric Design Pedestrian route type, (4) Commercial, C4: Urban
Controls: Freight route type, (5) General, C5: Urban Center,
• Design or Target Transit use C6: Urban Core
Speed considerations, (6)
• Design Traffic Complete streets and
Volumes Main Street highways
• Design LOS
• Road User
Attributes

56
No. NCHRP 839 WSDOT FDOT MnDOT ODOT Mass DOT
Ma Apply the • Select design controls • The design controls Fundamental Design Controls Design control and criteria: Basic design controls:
ss6 Appropriate compatible with the addressed in this manual (from PBPD) • Design vehicles: the number and • Roadway Context (MassHighway
Geometric Design context (Chapter 1103). include: • Level of Service: PLOS, type of trucks, functional classification PDDG Section 3.2)
Process and Criteria • Design controls create • Context Classification: BLOS, and TPLOS. of the highway, freight route • Roadway Users (Section 3.3):
significant boundaries Determines key design • Design Speed designation, and the effect on other Pedestrian, bicyclists, and drive
and have significant criteria elements for • Design Vehicle modes including pedestrians and • Transportation Demand (Section 3.4):
influence on design. arterials and collectors. bicycles, should all be considered. Design year, volume and composition of
WSDOT uses five • Functional classification Other design parameters: • Design speed: The selection of a the demand,
primary design controls: • Level of service • Functional classification design speed is dependent on traffic • Measures of Effectiveness (Section
(1) Design Year, (2) • Traffic and Design (MnDOT RDM p. 2-5(1)) volume, geographic characteristics, 3.5): facility condition, safety, mode
Modal Priority, (3) Year: Satisfy capacity • Types of highways functional classification, number of choice, network connectivity, level of
Access Control, (4) needs at an acceptable level (MnDOT RDM p. 2-5(3)) travel lanes, 85th percentile speed, service
Design Speed, and (5) of service through the • Traffic characteristics: (1) roadway environment, adjacent land • Speed (Section 3.6): selecting vehicle,
Terrain Classification design year. Volume, (2) Direction, (3) use, and type of project . pedestrian, and bicycle design speed
• Access Management: Distribution, (4) Composition, • Access management: Good access • Sight Distance (Section 3.7):
Regulation of access is and (5) Traffic flow. management will reduce the overall recognizing sight distance for motor
necessary to preserve the (MnDOT RDM p. 2-5(5)) number of crashes and increase the vehicles, bicyclists, and pedestrians
functional integrity of the • Speed (MnDOT RDM p. 2- highway’s capacity
State Highway System and 5(7)) • Traffic Characteristics: Four major
to promote the safe and • Capacity (MnDOT RDM p. components affect traffic
efficient movement of 2-5(8)) characteristics: (1) Vehicles, (2)
people and goods within • Sight Distance: Stopping Facility character and functional
the state. sight distance, Passing sight requirements, (3) Drivers/users, and
• Design Speed: (1) High: distance, Decision sight (4) Traffic demand (ODOT BUD)
50 mph and greater, (2) distance should be considered.
Low: 45 mph and less, and (MnDOT RDM p. 2-5(11))
(3) Very Low: 35 mph and • Terrain: (1) level, (2)
less. rolling, and (3) mountainous.
• Design Vehicle: The (MnDOT RDM p. 2-5(15))
largest vehicle that is • Crash data: Review crash
accommodated without history within project limits
encroachment on to curbs (MnDOT RDM p. 2-5(16))
(when present) or into
adjacent travel lanes.
(FDM)

57
No. NCHRP 839 WSDOT FDOT MnDOT ODOT Mass DOT
7 Designing the • Formulate and evaluate • Establish geometry, Use value engineering Work with different project team Document all considered alternatives
Geometric potential alternatives that grades, and cross sections principles to design members to refine the selected
Alternatives resolve the baseline need and evaluate alternatives alternatives: alternative design (ODOT HDM)
• Assemble an for the selected context • Use creative thinking to
Inclusive and and design controls speculate on alternatives that
Interdisciplinary (Chapter 1104). can provide the required
Team • Alternative Solution functions.
• Focus on and Formulation • Evaluate the best and lowest
Address the Need or • Alternative Solution life-cycle cost alternatives.
Solve the Evaluation • Develop acceptable
Problem(s) Within alternatives into fully
the Context supported recommendations.
Conditions and • Present/formally report all
Constraints recommendations to
management for review,
approval, and implementation.
(MnDOT RDM p. 2-4(2))
8 Design Decision Select design elements ✓ Document design decisions Document design decisions Document design decisions
Making and that will be included in
Documentation the alternatives (Chapter
• Independent 1105).
Quality and Risk
Management
Processes
9 Transition to Determine design ✓ ✓ ✓ ✓
Preliminary and element dimensions
Final Engineering consistent with
performance needs,
context, and design
controls (Chapter 1106).
10 Agency Operations
and Maintenance
Database Assembly
11 Continuous
Monitoring and
Feedback to Agency
Processes and
Database

58
Table 6: Summary of design processes

59
Table 7: Context classification comparison across jurisdictions

60
Table 8: Design control comparison across jurisdictions

61
After reviewing the design manuals and handbooks of the multimodal practice leaders, we

also worked to identify documents that address performance measures and data sources

recommended for project development. Two documents stood out. First is the ODOT

Blueprint for Urban Design which includes a table that lists some performance measures that

are useful for project evaluation across different modes, shown in Table 9.

Second, MassDOT’s guidelines on Complete Streets also established a set of performance

measures used for the development of Complete Street projects. These performance measures

are listed by mode in Table 10.

62
Table 9: Example of ODOT’s project-level performance measures by mode (ODOT, 2020b)

63
Table 10:List of MassDOT’s complete streets performance measures by mode (Lovas et al.,
2015)

64
2.3.5 Conclusions

Our interviews with experts and a review of design documents and guidelines from four

leading state transportation departments along with recent reports from NCHRP offer a

snapshot of the “state of the practice” for multimodal design. Broadly speaking, relative to

the past, best practice now eschews imposing a “one size, fits all” solution that focuses on

optimizing for automobile travel time. Instead, multimodal practice leaders favor solutions

tailored to a particular geographic and social context that balance the transportation interests

of car drivers, truck drivers, pedestrians, cyclists, and transit users. Best practice solutions

also consider other important social goals such as increased safety and reduced noise, air, and

water pollution. This shift in emphasis from one performance measure (level of service as a

proxy for travel time and reliability) to many performance measures adds complexity and

time to project design.

To address this complexity, modern best practice engages a broader range of stakeholders in

project design, considers a wider range of alternatives, and allows for more design flexibility

to match a solution to its specific context and potentially deliver more value for the project

dollar. Ideally, a multimodal design process employs a suite of performance measures that

address the competing stakeholder interests involved in a solution and allow them to see the

tradeoffs among competing goals. Unfortunately, the state of the practice has yet to develop

clear guidance on the appropriate set of robust and cost-effective performance measures that

a project designer should apply to a project of a given type. The development of standard

approaches to the use of multimodal performance measures represents the growing edge of

multimodal design practice

65
3 MisplacedWheels, A crowdsourcing tool for gathering
micromobility parking infraction data: the results of two
parking audit case studies in Seattle and Portland
Acknowledgement: This project was funded by C2SMART, and the work presented in this

chapter is a continuation of the paper: Pathak, C., Arabkhedri, B., & MacKenzie, D. (2021).

Crowd-sourcing Micro-mobility Parking Violation Reporting – User Interface Design

Motivation and Analytical Opportunities from Data Collected. Transportation Research

Board 2021 Annual Meeting, Washington, D.C.

3.1 Introduction and background

Micromobility services (i.e., an umbrella term for shared e-scooters and e-bikes) have

become increasingly adopted across various cities in the U.S. and throughout the world.

Micromobility companies may operate only bike, e-bikes, or e-scooter or operate a variety of

them. Micromobiloty services could further be categorized into two groups: dock-based

(meaning that there are several docks or hubs scattered around the city and all trips have to be

ended only at those locations) and dockless (meaning the vehicles can be left almost

anywhere and, on any sidewalk as long as they are parked correctly). These companies

normally have a smartphone app that allows the user to unlock a bike or scooter through

scanning a QR code on the devices.

Micromobility companies show a promising future having replaced short distance car trips

and provided first and last mile solutions for transit users (INRIX, 2019). Nonetheless, the

popularity that these vehicles have gained, while beneficial for increasing mobility, has

brought upon safety, equity, accessibility, and parking issues that all need to be addressed.

With the advent of dockless micromobility, there is not a requirement for users to end their

trips in a bike or scooter hub, and they can end their trip almost anywhere as long as they

66
abide by the parking regulations set forth by the cities. However, anecdotal observations as

well as peer reviewed research have found that the amount of micromobility parking

infractions across various jurisdictions is non-trivial and needs serious addressing. A

literature review of past studies (which consisted of both academic articles and reports by city

agencies) regarding micromobility parking violations found that the percentage of vehicles

which were not compliant with the rules ranged from 8% to 32.5% (median=16.1%) (Brown

et al., 2020). Furthermore, the report found that the percentage of vehicles blocking ADA

access ranged as high as 2.8% of the audited fleet (Brown et al., 2020). This shows that many

residents are potentially not fully aware of parking regulations in their cities which leads to

this high number of infractions (Brown et al., 2020).

Different jurisdictions and cities also have different rules from one another, which

complicates matters particularly for those living in two neighbor jurisdictions or using these

services in two different cities. Further, the rules could be as specific as “to not reduce the

sidewalk pedestrian zone to under 6 ft” to as broad as “to not park on vegetation and plants”.

A study which analyzed twelve policy dimensions (e.g. speed limit, parking rules, no. of

operators, etc.) across 10 peer cities to Nashville, did however find numerous areas of

consistency among the different policies (Janssen et al., 2020). This study complements prior

work by focusing specifically on parking policies, systematically comparing parking rules

across multiple cities in the US.

When the users cause parking issues, it is normally the companies’ responsibility to resolve

them per the permits and contracts that they sign with the cities. However, these infractions

may not get reported in the first place if there is not an easy tool for reporting. Jurisdictions

sometimes do not have an effective way to get reports on these parking violations and to pass

them on to the micromobility companies to resolve. One the other hand, there are

67
jurisdictions which use services such as a 311 service line, online reporting, or emailing the

micromobility coordinator of the department, to get informed on these issues.

We suggest a solution by developing a web application called MisplacedWheels to

crowdsource parking infraction data from residents and to also provide a tool for city officials

or contractors to conduct audits of the micromobility companies. The researchers used

anecdotal data from interviews with officials from five jurisdictions (including cities, an

unincorporated county, and one large university campus), to understand what the needs of

these various authorities are when they want to get reports on micromobility parking

infractions. The researchers also identified the parking rules of seven west-coast and one east-

coast cities (with different populations, geographies, and other attributes) by studying the

permits that these various cities had with the micromobility companies. They suggested a

system to rank the parking infractions from most to least severe based on a score of 1 to 10

(with 10 being the most severe and having higher priority). A “no infraction” option with

severity zero was also added for cases where the vehicle was parked correctly since the tool

was also purposed for conducting data audits and marking no-infraction cases. Once the app

was developed, the researchers conducted two parking audit case studies in the cities of

Portland and Seattle. The data from the case studies was very briefly analyzed in Pathak et al.

(2021). However, the paper did not include a thorough analysis of the data as it was primarily

a motivation for the development and introduction of the app. This chapter, in addition to

introducing some of the motivations behind the study and briefly discussing some jurisdiction

parking rules, further includes a thorough analysis of the collected data.

The data is cleaned and merged with other datasets such as city shapefiles, census tract

demographics, socioeconomic attributes, and built environment data such as number of bike

racks and bus stops. The data is aggregated on a census tract level and is fed to a linear

68
regression model to find any relationships between independent variables such as median

income, number of bike racks and bus stops, and population density in a census tract, and the

percentage of parking violations in each census tract. Other models also analyzed the

relationships between the named attributes and the percentage of high severity violations

(severity >= 7), the number of violations, and number of reports from each census tract (to

control for any biases).

This chapter is organized as follows. First, a brief summary of the interviews with city

officials is mentioned by highlighting essential attributes requested in the app before moving

to an analysis of the parking infraction rules in eight cities. Then the data audit sampling plan

for the case studies is described in the data collection section. Data cleaning methods are

mentioned in the next section, and then we move to spatial analysis of the data with

numerous maps to understand the geographical distribution and prepare for the modeling

section. Finally, we discuss several models built to find relationships between census tract

demographics and/or built environment characteristics with the parking infractions and

highlight the results. Lastly, some conclusions are drawn and future directions are discussed.

3.2 Benchmarking micromobility infraction reporting tools and results

from interviews with five jurisdictions

Before conducting interviews, 13 U.S. cities were benchmarked in the fall of 2019 to improve

understanding of how each of them handled micromobility parking infractions in their

jurisdictions. These cities were chosen as they all had active micromobility programs. There

was no systematic method in choosing these cities; but, there was a slightly higher focus on

Northwestern cities. The cities chosen were (in alphabetical order): Atlanta, Austin, Bellevue,

Chicago, Dallas, Houston, Los Angeles, New York, Portland, San Francisco, Seattle,

Spokane, and Washington D.C. Table 11 shows the results of this benchmarking.

69
In this not very large sample, it was found that nine cities suggested to directly contact the

bicycle/scooter companies to report infraction cases, five asked to call a local 3-1-1 number,

three asked to report to the city’s transportation department through email or calling, two

mentioned to use a web portal to report the issue, and lastly, four guided users to report cases

with the city’s mobile app. The apps for Los Angeles and Dallas were namely myLA311 and

Dallas311, and while Los Angeles’s app did have a specific section for reporting

micromobility infractions, Dallas’s app only had a general section which could be used for

micromobility as well. Seattle’s FindItFixIt app and Bellevue’s app both had built-in sections

for reporting micromobility infractions. Nevertheless, Table 11 shows that even though some

cities offer multiple solutions, many of them still rely on users contacting the companies

directly; in fact, seven out of the nine that suggest directly contacting the companies, identify

this method as the quickest and most efficient way to resolve issues.

After the benchmarking, the authors conducted interviews of five jurisdictions, including

three cities, one unincorporated county, and one large university campus, which all had their

own permits with either one or various micromobility companies. These interviews were

conducted in the winter of 2020 before the COVID-19 pandemic. The names of these

jurisdictions are kept anonymous; however, we will mention that all of them were based in

the Pacific Northwest Region of the United States. These jurisdictions were asked about a set

of key data and features that they deem as necessary information to be collected from

micromobility parking infraction reports, particularly, information that would be useful both

for themselves and for the bikeshare and scootershare companies in order to better manage

parking violations.

Even though they all used different ways for handling infraction reports, they all mentioned

some key features which need to be collected. The results of these interviews are shown in

70
Table 12. The table does not specify which of the jurisdictions were cities, an unincorporated

county, and a university campus; they are brought in no particular order to keep their

anonymity. Admittedly, this is a small sample size and did not serve as an ultimate guide to

what we should and should not collect in an infraction reporting app; nevertheless, it did help

us narrow down some key features that would later be implemented in the application.

Geolocation, photo, QR code, description of the problem, and priority of the violation were

all features that were pointed out by the jurisdictions. Other interesting notes from the

interview included the use of QR codes to verify whether the same bike was moved to

another location leading to another violation or if an issue is reported several times, making it

a high priority. The “level of frustration” with the infraction was also mentioned as a separate

attribute from the “priority” of the parking infraction. The level of frustration is inherently

different from the priority since, for instance, the presence of a scooter on one’s private

property may highly frustrate them while the cities and vendors may not see that as a high

priority and aim at resolving more urgent issues such as ADA access blockage.

Another important note brought up was that in areas with neighboring jurisdictions having

different operating companies, people may sometimes not be aware of the exact boundaries

and move bikes and scooters from one jurisdiction to another. Further, in a number of cases,

parking complaints which are the responsibility of one jurisdiction entity end up getting

reported to another. This situation is further aggravated as sometimes, 1) a certain company

may be operating in one and not the other; and 2) different jurisdictions may have different

regulations in their permits (i.e., a situation that may be deemed as an infraction in one, may

not necessarily be an infraction in the other). As a result, these jurisdictions pointed out that a

singular reporting app which is geo-sensitive and would automatically provide a list of

company names and infractions based on the location would be a useful tool for them in

71
resolving these (not so very often) issues with neighbor jurisdictions.

Table 11: Method(s) used by 13 U.S. cities for resolving micromobility parking infractions

Table 12: The results of interviews with five jurisdictions specifying key and good-to-have
features that they require a micromobility infraction reporter tool to incorporate

72
3.3 Review of parking infraction rules in various cities

This section reviews parking regulations across seven different, various-sized cities. Six of

the jurisdictions are west-coast cities with five being in the pacific northwest (Seattle,

Redmond, and Spokane in Washington, Portland, OR, and Boise, ID) and two in California

(San Francisco and Los Angeles), and one on the east-coast (Washington D.C.). These cities

were chosen as we initially had plans to conduct some scale of data collection in all of them

and needed to implement the city-specific rules into our developed app for them. As we had

collected all their rules, we sought this as an opportunity to group these rules into several

categories in this section and understand to what degree these rules are similar or different.

The results can be seen in Table 13. Nine categories were defined and are as follows: 1)

Sidewalk related, 2) ADA (rules in accordance to the Americans with Disabilities Act), 3)

Blocking access, 4) Fastened to a street feature, 5) Rules related to the curb zone, 6) Rules

related to emergency and utility facilities, 7) General no-parking areas, 8) Unauthorized

parking areas, 9) Rules specific to a particular city. Our findings showed that parking

regulations were to some extent similar for these different cities while some unique infraction

rules were observed for cities like Portland, San Francisco, and Washington D.C.

We will start those that are similar. All cities had a rule which does not allow to reduce the

pedestrian clear zone of the sidewalk to under five or six feet. Additionally, some ban parking

in a sidewalk where the furniture zone is under three feet. They all had a rule that the scooters

should not be toppled and should be parked upright (we categorized this as sidewalk-related).

All cities banned vehicles from blocking access to ramps, doorways, entryways, travel lanes,

bicycle lanes, and dedicated transit lanes (ADA related). Further, all of them disallow parking

in a transit platform. They all do not allow bikes or scooters to be parked in someone’s

73
private property and most of them ban these vehicles from being parked in planted areas or

landscaping.

Most of them do not allow scooters to be placed next to street features or fastened to them.

While some cities such as Portland and Washington D.C. do not allow scooters to be fastened

to bike corrals, in D.C. it is required that shared bikes be locked to legal street infrastructure.

Most cities have some sort of restriction for not parking in a geofence-restricted parking zone

and restrictions for bikes or scooters being parked in one location for more than certain

consecutive days (1, 5 or 7). However, Portland and Seattle have more specific rules such as

not allowing scooters to be gathered in a cluster with more than 10 ft. of length and not

having 15 or more devices in a 1000 ft. block face.

These infraction rules were also given a severity level from 1 to 10, with 1 being the least

severe and having less priority, and 10 being the most severe and needing immediate

resolving. Cities can adjust these severities on their own based on what they identify as

higher priority for their operations. A list of the separate rules for each of the eight cities is

also available in Appendix A.

74
Table 13: List of parking infractions for different cities grouped into 9 categories (SEA = Seattle, PDX = Portland, SF = San Francisco, LA = Los Angeles, DC
= Washington D.C.)

75
3.4 App development

Our research team developed a mobile web application called MisplacedWheels throughout

the winter and spring of 2020. Since it is a web-app, it means that it operates on any browser

on almost any phone as it does not depend on the OS of the device. The user interface of the

app has three essential steps for reporting a mis-parked vehicle: location, classification, and

identification. In the first step, the user can pin the location of the mis-parked vehicle on a

map or use the smartphone’s GPS to find the location. In the second step, the user classifies

the type of infraction by selecting one or several items from a list of infraction rules and

selects the company of the vehicle. In the last step, the user can identify a specific vehicle by

scanning the QR code. Further, step 2 also enables the user to upload a photo of the mis-

parked vehicle. The details of the app development are mentioned in Pathak et al. (2021).

Some screenshots of the app are brought in Figure 9.

Figure 9: From left to right (a) location page, (b) classification page, and (c) identification
page of MisplacedWheels.com (Pathak et al., 2021)

76
The list of infractions and micromobility companies are sensitive to the city. The parking

infractions and companies of eight cities are implemented in the app all previously named in

Table 13, thus the rules in the app are in accordance with the government’s requirements for

parking in those cities. The order of the list of infractions is sorted by severity with higher

infractions shown on top and lower ones shown in the bottom.

3.5 Data audit conceptualization and sampling plan

One purpose of our research team after developing the app was to partner with other cities

and offer them this service so that it could enhance their micromobility management system

and serve as a data solution for their current lack of information regarding micromobility mis-

parking. From our viewpoint, such a resource could not only help them resolve parking issues

quickly and efficiently but also to help understand parking infraction hotspots in the long-

term. However, due to the spread of the COVID-19 pandemic in early 2020, micromobility

ridership dropped severely per the stay-at-home orders and many micromobility operations

were disrupted. Thus, our initial goal of partnering with cities and collecting data through

large volumes of crowdsourcing was not accomplished. Nevertheless, we sought the

opportunity to conduct a data audit ourselves to be able to answer some of our research

questions.

A data audit of the parking situation was performed by the research team in the cities of

Seattle and Portland. Seattle was chosen due to the researcher living in the area and easy

access for the author to conduct a data audit. Portland was chosen due to its proximity to

Seattle and also because it was representative of more micromobility companies (five

companies at that time). Data was collected for four days in Portland from July 18th to 21st

of 2020, and for two days in Seattle from July 22nd to 23rd of 2020.

77
Seattle, at the time, had only about 200 electric bicycles in its boundaries. All the e-bikes

were for the company JUMP. This was before the introduction of several new scooter

companies that arrived in the city around September and October of 2020 (i.e., LINK,

Wheels, and Lime, along with the previous JUMP bikes). Portland, on the other hand, was a

bigger hub for micromobility vehicles at the time, compared to Seattle, and all of its

micromobility companies operated e-scooters. In terms of sheer number of micromobility

vehicles, there were around six times more e-scooters in Portland at that time compared to

Seattle's roughly 180 e-bikes.

To locate and find the number of scooters available across both cities before and during the

audit, an app named ScooterMap was identified and used by the authors as a primary source

(Pontis, n.d.). This app contained information on all scooters and bikes that show up on the

map by pulling data from the micromobility companies' APIs. This app was used both prior

to the study, for counting the number of micromobility vehicles in each city and preparing a

sampling plan, as well as during the study, for finding the vehicles' locations. Figure 10

shows a screenshot of the Scooter Map app and its user interface. The pins on the map

display where the micromobility vehicles are parked, with the logo of the company being

shown on the left-side of the pin and the type of the micromobility vehicle (bike or scooter)

on the right-side. The colors of the pins indicate the battery level of the vehicles with red

being the lowest, green being the highest, and yellow in between (the thresholds are not

known). As of June 2021, when this thesis was written, the app is no longer available.

For the City of Seattle, the author collected data by doing an audit of all scooters that showed

on the map. As there were only about 180 e-bike available in the city (according to

ScooterMap), it was easily possible to audit all the bikes in two-days’ time. For the City of

Portland, however, a sampling plan was prepared to sample roughly half of the scooters there

78
since it had a much larger number compared to Seattle. The sampling method for Portland is

mentioned in the following subsection.

Figure 10: The user interface of Scooter Map (Lekach, n.d.)

3.5.1 Sampling method for Portland

There were about 1160 scooters in the city of Portland according to Scooter Map (Pontis,

n.d.) on 7/15/2020, three days before the data collection. As this is a relatively large number,

a census of all the micromobility vehicles was not possible to carry out, and thus we had to

sample a limited number. Therefore, a data sampling plan was gathered in preparation for the

data collection.

We wanted to sample about half of the total number of scooters in the city, meaning around

600 scooters. Having done smaller-scale tests of the app before, we established that each

report submission in the app would roughly take 30 seconds, an attribute that can also

become faster as the user gets comfortable with the app. Also, traveling between every two

79
scooters was assumed to take around 3.5 minutes, accounting for the walking and driving of

the auditor, within and between neighborhoods. As a result, it was assumed that 60/4=15

scooter reports can be submitted per hour. By taking into account 10 hours of auditing per

day (excluding lunchtime and rests), this would add up to 150 scooters per day. Thus the

sample of 600 scooters could be achieved in four days of data collection.

Cluster sampling was proposed as the sampling method. In cluster sampling, the entire data

frame is divided into several disjoint clusters. Zip codes, census tracts, and census block

groups are all popular for achieving different scales in cluster sampling. However, based on

our scale of data collection (citywide scale), zip codes were ultimately chosen as a suitable

disjoint cluster for this study.

According to the City of Portland’s official website, Portland has 32 zip codes. However, a

number of these zip codes are only partly within city boundaries (i.e., a very large portion of

the zip code’s area is outside city boundaries, for example, more than 90% of the area) or are

suburban areas with scooters being virtually or entirely non-existent in them according to

Scooter Map. After excluding such areas and using Scooter Map to find zip codes with at

least 20 scooters, a total of 16 zip codes were identified as viable clusters, shown in Figure

11.

Having divided the entire frame into 16 disjoint viable zip code clusters, we then wanted to

select a random sample of eight clusters (half of the total cluster). Using Scooter Map's real-

time data, we had a rough estimate of the number of scooters available for rent in each of the

zip codes. It is said rough, and not exact, because scooters may constantly change locations as

scooter users move them from one neighborhood to another; nevertheless, it was assumed

that chronological changes in the number of scooters per zip code are not major. We wanted

80
to ensure that the sum of scooters in the eight random clusters would result in our overall

achievable number of 600 scooters for the four-day data collection.

The sampling process was (admittedly) not entirely random due to not accepting the first

random few samples since they did not result in a number of around 600. However, it only

took a few random tries until the setup of the eight sampled corridors reached an overall

number of around 600. This meant a slightly higher concentration in urban and central areas

to achieve higher scooter numbers. The selected clusters are shown in Figure 12. Further,

Figure 13 shows the actual collected data overlaid on the same map to show that the data

collection was in accordance with the sampling plan.

In addition to those scooters which showed up on Scooter Map, several other scooters that

were not available for rent in the app were also identified when the auditor was walking in

various neighborhoods. These scooters did not show on scooter map because they were

disconnected from the API; nevertheless, they were collected in our data collection leading to

the larger number of samples than what we expected (around 690 instead of 600). In addition,

there were a near-total of 20 scooters (out of the 690 scooters audited) that were in a

neighboring zip codes but very close to the zip code boundary. Even though these were not

within the boundaries of the eight zip codes they were collected by the researcher. Lastly, the

reason why it took four days to collect 690 data points in Portland and two days to collect

nearly 160 data points in Seattle goes back to the fact that scooters in Seattle were more

geographically dispersed. The data audit involved less walking and mostly driving between

neighborhoods, causing a higher travel time between every two scooters.

81
Figure 11: Zip codes with at least 20 scooters shown with a light blue shade. Thick border
depicts Portland’s boundaries.

Figure 12: Eight randomly sampled zip codes, shown in a darker shade of blue, that will be
used as the disjoint clusters for the study

82
 Figure 13: The actual collected data overlaid on the zip code map after the data collection.
The map shows that all collected points are in accordance with the sampling plan as they are
within the dark blue shade.

3.6 Data cleaning, post-processing, and summary statistics

After data collection, the data was post-processed in R language (R Core Team, 2021) to

achieve the following three goals: 1) to clean any duplicate or unrelated data, 2) create

summary statistics, 3) to spatially plot and analyze the data, 4) join the dataset with other

resources such as socioeconomic and demographic info from the census and data from the

built environment such as number of bus stops, bike racks, etc. to model the data.

3.6.1 Data cleaning and post-processing

There was not an intense amount of cleaning needed after the data collection. Base R libraries

and the dplyr package were used for all the cleaning purposes. First, as was seen in Figure 13,

there were two points in the eastside 97233 zip code that laid slightly outside of Portland's

83
boundaries in the City of Gresham. Those two points were yet included for Portland and not

removed even though they were in a neighboring city. Another data cleaning task was to

ensure that all reports had a single company listed as their company label. The auditor made

sure, during the data collection, that he submits an individual report for each scooter, even if

they were only several inches away from each other. Nevertheless, there was one case where

two scooters from different companies were reported in one report with two company labels.

Thus, two individual reports were created for that single report during post-processing, and

the original report was removed. Lastly, for the case of Seattle, the University of Washington

(UW) had a different set of rules from the city and had its own permit with JUMP and ran

autonomously from Seattle's jurisdiction. We had also incorporated UW’s rules within its

boundaries and geo-referenced it into the app. Thus, during the two-day Seattle data

collection, there were 12 reports generated from within UW's jurisdiction that are not brought

in this thesis. With those being removed, 156 reports were submitted all under Seattle's

jurisdiction, and 690 reports for Portland. Other than these named tasks, no further data

cleaning was needed.

Figure 14 shows the total number of reports from each city by breaking them into violation

and non-violation categories. It was found that out of the 690 vehicles audited in Portland,

292 were in violation of the city's rules (0.42 violation rate), and out of the 156 vehicles

audited in Seattle, 95 cases of violations were seen (0.61 violation rate). Portland seems to be

better off compared to Seattle in terms of the violation rate. A two-tailed hypothesis test was

performed to see if there is a significant difference in the portion of violations between the

two cities. The null hypothesis was that there is no difference between the two cities, and the

alternative was that there is a difference. Before testing, we looked over the conditions for

inference meaning to make sure both samples were random, normal, and independent. The

sample from Portland was taken through random cluster sampling so it does somewhat meet

84
the condition of randomness, however, for Seattle, the entire population of scooters was

studied. Even though this means the conditions of inference are not met, the hypothesis test

was still performed. α=0.05 was assumed.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

According to Equations 1 to 8, it was found the null hypothesis can be rejected, and the

difference between the rate of parking violations in Seattle and Portland is significant.

However, again, this test may not entirely meet the conditions of inference as the data for

Seattle is not a sample and is the entire number of shared e-bikes in Seattle.

85
Figure 14: The number of violation and non-violation reports stacked on top of each other for
Portland and Seattle. The height of the bar is the total number of reports for each city.

3.6.2 Summary statistics

Below are some summary statistics of the data to help understand them better. Figure 15 and

16 show the number of vehicles audited for each company, for Seattle and Portland,

respectively. As it was mentioned, at the time of the data collection, only JUMP e-bikes had a

permit and were available in the City of Seattle; however, there were also two Lime e-

scooters observed in the audit. It was perceived that the scooters were moved from a nearby

county to Seattle. Regardless, they were included in the count of the data. For Portland, Spin

had the highest number of e-scooters in the city, followed by Lime, Bird, Bolt, and Razor.

Figures 17 and 18 then show the number of violations next to the total number of vehicles

audited for each of the companies, in Seattle and Portland, respectively. Furthermore, Figures

19 and 20 display the portion of different severities for each of the companies, in Seattle and

Portland, respectively. Severity 0 means no violation, and severities 1 through 10 all mean

some sort of violations with 1 being the least serious, such as the vehicle being parked on

86
vegetation, and 10 being the most serious, such as vehicle reducing sidewalk width to under 6

ft. It can be seen that in Portland all of the companies have more than 50% of scooters

without any violation (severity = 0), with Bird doing exceptionally well by having more than

75% of scooters not violating the rules. It was anecdotally observed during the data collection

that Bird scooters were maintained better than others as they were mostly grouped together

on the furniture zone of the sidewalk probably by a company worker, thus resulting in a

lower violation rate. Regardless, all companies had more than 10% of scooters with a severity

of 10 (meaning either blocking the sidewalk or causing an access issue for people with

disabilities) with most of them having a rate near 20% or 25%.

Lastly, Figures 21 and 22, demonstrate the distribution of the violation severities (severity =

0, meaning non-violation, was excluded), for Seattle and Portland. Histograms and density

plots were used for these figures. While there was a slight element of subjectivity blended

into the severity rankings, the distribution plots are yet alarming due to their high numbers of

severity 10 (mostly being reducing the sidewalk pedestrian clear zone to under 6 ft or causing

an access issue for the disabled). Seattle, having a violation rate of 0.62, higher than that of

Portland, has 35% of violations with severity of 1. Both cities have roughly 45% of violations

with severity of 10. Seattle has a mean severity rate of 5.5 while Portland's is around 7.1. A

hypothesis test was not performed to see if the difference is significant due to the element of

subjectivity that was discussed.

After looking at the summary statistics, some spatial analysis of the data was conducted along

with gathering census data for modeling data.

87
Figure 15: Count of micromobility vehicles audited in Seattle per scooter company operating
there, color coded by the company's colors

Figure 16: Count of micromobility vehicles audited in Portland per scooter company
operating there, color coded by the company's colors

88
Figure 17: Number of vehicles audited, plotted next to the number of violations for each
micromobility company in Seattle

Figure 18: Number of vehicles audited, plotted next to the number of violations for each
micromobility company in Portland

89
Figure 19: Portion of different severities from 0 (meaning no violation) to 10 (severities 1
through 10 all mean some sort of violations) for Seattle

Figure 20: Portion of different severities from 0 to 10 for Portland

90
Figure 21: Histogram and density plot showing the severity distribution of violating
micromobility vehicles in Seattle (severity 0 is not included)

Figure 22: Histogram and density plot showing the severity distribution of violating
micromobility vehicles in Portland (severity 0 is not included)

91
3.7 Spatial analysis

In this section, multiple attributes of the data are plotted across a variety of maps to better

understand the distribution of the data. The severities of the violations are also shown through

a heatmap, and the infraction situation across different census tracts is shown using an

interactive plot. The distribution of transit stops and bike racks across the city are also

overlayed on the infraction reports data to visually see if there are any correlations between

these attributes.

Even though an interactive spatial analysis tool for the data is already available online at

Misparkrepo.com (Pathak et al., 2020), the micromobility infraction manager tool which was

developed by the team, this section is an effort to go beyond the tools already offered in

Misparkrepo. We aim to particularly point out map aesthetics that may help in identifying

infraction hotspots more effectively as well as another interactive tool which allows for

aggregate neighborhood comparisons rather than seeing single infraction report spots.

The data was plotted in R using “tmap” (Tennekes, 2018) and “ggmap” (Kahle & Wickham,

2013) libraries. First, Figure 23, shows a heatmap of the severities for Portland while Figure

24 shows a zoomed-in version for Portland with a particular focus in Downtown and Central

Portland areas. Figure 25, shows the same map type for Seattle. The choice of color for these

figures is primarily important as it highlights those infractions with higher severity (in purple)

while displaying the lower priority infractions in yellow. Further, an alpha parameter of 0.3 is

set for the glyphs (circles) to allow for better transparency and visualization for points that

overlay each other. This particularly helps to better monitor denser areas such as the case in

Downtown Portland (Figure 24) where many reports are generated in very close blocks.

Additionally, the glyph size also varies for different severities to bring more concentration to

areas with higher severity. Many of these aesthetic points come from points discussed in a

92
data visualization course by Chris Adolph and from Edward Tufte’s The Visual Display of

Quantitative Information principles (Tufte, 1985).

Another data exploration opportunity sought was the visual display of aggregated census

tract-level infraction rates. Infraction rate is defined as the number of vehicles in a census

tract not compliant with rules, divided by the total number of audited vehicles in that census

tract (i.e., number of violating vehicles / total number of audited vehicles, for each census

tract). Furthermore, high severity infraction rate is defined as the number of vehicles in a

census tract with a severity of higher than seven, divided by the total number of audited

vehicles in that census tract.

These two metrics would allow for a comparison of how various census tracts are performing

compared to one another. Figure 26 and 27 show the infractions rates and high severity

infraction rates, respectively, for various census tracts in Seattle where data were collected

from. Figure 28 and 29 show the same respective features for Portland. Figures 30 and 31

show the same but they are zoomed in to show Downtown and Central Portland areas as the

number of reports are denser there. An interactive version of these maps is available for

Seattle and Portland for parking infraction rates (Arabkhedri, 2021b) and high severity

parking infraction rates (Arabkhedri, 2021a) per census tract.

93
Figure 23: Distribution and severity of parking infraction reports in the Portland case study

94
Figure 24: Distribution and severity of parking infraction reports zoomed in on Downtown
and Central Portland areas for better visualization

95
Figure 25: Distribution and severity of parking infraction reports in the Seattle case study

96
Figure 26: Micromobility parking infraction rates per census tract for the City of Seattle
shown with different colors (purple for highest rate bin and yellow for lowest rate bin).
Interactive version available at: https://rpubs.com/bornakhedri/micromobility-parking-
infraction-rate

97
Figure 27: Micromobility high severity parking infraction rates per census tract for the City
of Seattle shown with different colors (purple for highest rate bin and yellow for lowest rate
bin). High severity means only those with a severity greater than or equal to 7. Interactive
version available at: https://rpubs.com/bornakhedri/micromobility-high-severity-parking-
infraction-rate

98
Figure 28: Micromobility parking infraction rates per census tract for the City of Portland
shown with different colors (purple for highest rate bin and yellow for lowest rate bin).
Interactive version available at: https://rpubs.com/bornakhedri/micromobility-parking-
infraction-rate

99
 Figure 29: Micromobility high severity parking infraction rates per census tract for the City
of Portland shown with different colors (purple for highest rate bin and yellow for lowest rate
bin). High severity means only those with a severity greater than or equal to 7. Interactive
version available at: https://rpubs.com/bornakhedri/micromobility-high-severity-parking-
infraction-rate

100
Figure 30: Zoomed in version for Figure 27, concentrating mainly on the Portland central
area.

101
 Figure 31: Zoomed in version for Figure 28, concentrating mainly on the Portland central
area.

3.8 Data modeling and results

For the data modeling section, census data was gathered through the "tidycensus" package

(K. Walker, 2020) in R, from the American Community Survey (ACS) 2019 5-year survey.

The variables gathered were: total population, 1st to 5th quartile mean incomes, top 5% mean

income, median income, total number of homeowners, total number of vehicles, percentage

of white population (to then calculate the minority rate), and median age. These were

102
gathered for every census tract where at least one report was submitted. Further, the area for

each of the census tracts was gathered using the "tigris" package in R (K. E. Walker, 2016).

From there on, some other variables were calculated by combining the gathered attributes so

far, i.e., percentage of homeowners (number of homeowners / population), minority ratio (1 -

white/total population), number of vehicles in a census tract divided by population, and

population density (total population / area in squared kilometers).

Data regarding built environment factors were sought as well. Two data attributes that made

sense for this analysis were the number of transit stops and the number of bike racks in each

census tract. Data for the bus stops in Seattle was collected from King County Metro's GTFS

database. Bike racks in Seattle were collected from the City of Seattle's open Geo Data

including bike racks owned and maintained by SDOT. Transit stops in Portland were

gathered from the TriMet developer resources. Lastly, Portland bike racks were collected

from Portland Maps Open Data, although they were not used for the modeling as Portland's

micromobility vehicles were all scooters and not bikes.

Eight Ordinary Least Squares (OLS) linear regression models were created from the cleaned

data, four for Portland and Seattle each. The final independent variables used were: census

tract median age, total population, mean income of the top 5%, median income, portion of

home owners, minority rate, average number of vehicles in census tract per person,

population density, number of transit stops in each census tract (for both Portland and

Seattle), and lastly, number of bike rack in each census tract (for Seattle only). The dependent

variable for the four models were: 1) portion of reports in a census tract that were a violation

(infraction rate), 2) total number of violations per census tract, 3) portion of reports in a

census tract that had a severity of equal to or higher than seven (high severity infraction rate),

and 4) total number of reports per census tract. The last dependent variable was particularly

103
modeled to control for any biases in the data collection and improve understanding of

whether the sheer number of reports generated in each census tract was correlated with any of

our independent variables.

The main advantage of an OLS model is its simplicity and that it is easy to interpret.

However, some of its disadvantages include being sensitive to outliers and that the test

statistic might be unreliable if the data is not normally distributed. An alternative would have

been the use of a Robust Linear Regression model as there were some outliers in the data. In

addition, a logistic regression could also be used to model for each audited vehicle which has

a binary label of violation or no violation, which may have kept more richness in the data

compared to when it is aggregated over each census tract for a linear regression model.

A stepwise regression method was used to ensure independent variables that have a high

correlation with the dependent variable stay in the equation and that insignificant variables

are omitted. The model would optimize itself to achieve the highest AIC by going back and

forth and dropping or adding independent variables. The downsides of this method are that it

is prone to overfitting as well as prone to spurious correlations (i.e., factors that are not

causally related and are correlated by coincidence or a third unseen factor). All models start

with zero variables, then add the most contributive predictors sequentially, and then remove

any variables that no longer provide an improvement in the model. The stepAIC function was

used from the MASS package (Venables & Ripley, 2002).

Results for the Seattle model are shown in Table 14. The linear regression model for Seattle

showed that the number of transit stops in a census tract is positively correlated with the

percentage of violations (violations divided by the total number of reports per each census

tract) in that same geography, and the number of bike racks is negatively correlated with the

percentage of violations. Per SDOT's policy, which is to encourage users to park their

104
bikeshare vehicles next to bike racks, we also expected fewer violations in areas with more

bike racks. The sheer number of violations in each census tract were also found to be

significantly correlated with the same two independent variables in the same manner. The

third model showed that percentage of high severity violations negatively correlates with the

population density and number of transit stops. The model for Portland, shown in Table 15,

did not illustrate as interesting results.

Table 14: Linear regression models for Seattle census tracts

Table 15: Linear regression models for Portland census tracts

105
3.9 Findings and conclusions

This work talked about the development of and the motivation behind a web-based

application for the purposes of crowdsourcing micromobility parking infraction data from

residents and for conducting data-driven parking audits of micromobility companies. The

work benchmarked the way 13 different cities ask citizens to report micromobility parking

infractions. It further conducted interviews of five different jurisdictions to ask about key

attributes needed for a micromobility parking infraction reporting app. Also, the work studied

policies regarding micromobility parking in eight U.S. cities while categorizing them into

nine different groups, underlining their similarities and differences. The work presented a

sampling plan used for conducting a parking data audit in the cities of Portland and Seattle.

Data collected from the two case studies were cleaned and briefly summarized through

various plots. Furthermore, the data was analyzed spatially to show the distributions and

severities of parking infractions across the two cities. An interactive visualization tool was

introduced to count the portion of parking infractions and high severity parking infractions on

an aggregate census tract level. Lastly, the data was modeled using an OLS linear regression

model to seek relationships between parking infraction and factors of the built environment

and socioeconomic and demographic factors.

We found that neighborhoods with more bike racks in Seattle had a less percentage of their

reports being parking infractions. Also, areas with more transit stops were found to have a

higher percentage of parking infractions. However, these results did not hold true for the data

from Portland. It is recommended that the same study be conducted in other cities to find

better relationships between the number of parking violations and other built environment

characteristics.

106
One point that this study was aiming to emphasize was the importance of crowdsourced data

in performance monitoring of programs in different cities. This app was designed to serve

both as an auditing tool for city representatives or contractors, but also for the general public

to report parking infractions in their neighborhoods. We also initiated some efforts of

coordinating with various cities to deploy this app as a state-of-the-art solution for reporting

parking infractions; however, impediments such as the COVID-19 pandemic and the

uncertainty for bike and scooter companies during that period, certainly did not help our

efforts.

Rooms for improvement and future research could include the use of logistic regression for

evaluating whether an infraction would happen or not based on variables such as distance to

the nearest transit stop or to the nearest bike rack (instead of looking at numeric independent

variable at aggregate level, we can look at individual reports with a dichotomous label of

violation or no violation). Furthermore, the use of other state-of-the-art machine learning

algorithms for predicting the violation labels based on census tract

demographic/socioeconomic factors and/or built environment information could be valuable.

Nevertheless, this tool can be used as a data solution for handling micromobility parking

infractions in the long run. The modeling and spatial sections of this study can be used as an

analysis framework for cities to evaluate how various micromobility companies are doing and

if they are abiding by the rules in their permits.

107
4 Conclusion
This work was an effort to improve the understanding of effective multimodal transportation

design processes and performance-based practical design decision making through

conducting a national-scale literature and document review. Chapter 2.2 synthesized the state

of the knowledge via reviewing the available academic literature, and Chapter 2.3.

synthesized the state of the practice by reviewing practical documents such as design manuals

and guidebooks from WSDOT, ODOT, FDOT, MnDOT, and MassDOT, as well as two

NCHRP reports. The work introduced key terminology, discussed multimodal performance

measures, compared the current design processes across various jurisdictions, and drew

conclusions on what jurisdictions should do to advance their multimodal design. The

contribution of this work is that it gathers literature from various sources and establishes that

both scholarly research and practical guidelines are suggesting relatively similar solutions to

advancing multimodal transportation, and that performance-based design and engineering

concepts should be adopted by practitioners to ensure access, safety, and mobility for all

modes.

The second section of the thesis focused on a new data-gathering tool and analysis framework

that could potentially help cities more effectively monitor the performance of micromobility

services (i.e., shared e-scooters and e-bikes). The use of a crowdsourcing application

developed by our research team could help cities monitor the performance of micromobility

vehicle programs. This tool can also be used for conducting data-driven parking audits of

bikes and scooters, on a neighborhood and/or a city level. Results from two parking audits in

the cities of Portland and Seattle showed that this tool can be used to spatially analyze the

data, create summary statistics to compare metrics for various companies, and help spatially

analyze the data. Several statistical models were also developed to seek for links between

parking violations and elements of the built environment and/or census tract socioeconomic

108
factors and demographics. This data collection tool and analysis framework can be used by

cities to conduct parking audits of micromobility companies, and to ensure that the

transportation network and mobility programs are operating successfully through

performance monitoring.

109
References
Arabkhedri, B. (2021a). Micromobility High Severity Parking Infraction Rate per Census

Tract—Seattle & Portland. RPubs. https://rpubs.com/bornakhedri/micromobility-

high-severity-parking-infraction-rate

Arabkhedri, B. (2021b). Micromobility Parking Infraction Rate per Census Tract—Seattle &

Portland. RPubs. https://rpubs.com/bornakhedri/micromobility-parking-infraction-

rate

Barbeau, S. J., Pham, M., Nieves, J. A., Bertini, R. L., University of South Florida, T., Center

for Transportation Equity, D. & D. (CTEDD), & Office of the Assistant Secretary for

Research and Technology. (2020). Improving Transportation Performance

Measurement via Open “Big Data” Systems – Phase 1 Transit (01727573; p. 37p).

https://ctedd.uta.edu/wp-

content/uploads/2019/07/ImprovingTransportationPerformance_Barbeau_CTEDD_20

19.pdf

Brown, A., Klein, N. J., Thigpen, C., & Williams, N. (2020). Impeding access: The frequency

and characteristics of improper scooter, bike, and car parking. Transportation

Research Interdisciplinary Perspectives, 4, 100099.

Cambridge Systematics, I. (2010). Measuring Transportation Network Performance

(01173400; Issue 664). Transportation Research Board.

http://www.trb.org/Publications/Blurbs/164141.aspx

Cambridge Systematics, Inc. (1999). Multimodal transportation: Development of a

performance-based planning process. NCHRP Research Results Digest.

https://apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=900

110
Dock, S., Westrom, R., Lee, K., & Cesme, B. (2017). Advancing Urban Multimodal

Transportation System Performance Management. Transportation Research Record:

Journal of the Transportation Research Board, 2646, pp 17-27.

Dowling, R. G., McLeod, D. S., Guttenplan, M., & Zegeer, J. D. (2002). Multimodal

Corridor Level-of-Service Analysis. Transportation Research Record: Journal of the

Transportation Research Board, 1802, 1–6.

Dowling, R. G., Reinke, D. B., Flannery, A., Ryus, P., Vandehey, M., Petritsch, T. A.,

Landis, B. W., Rouphail, N. M., & Bonneson, J. A. (2008). Multimodal Level of

Service Analysis for Urban Streets (NCHRP Report 616). ISSN 0077-5614.

https://trid.trb.org/View/870308

Elefteriadou, L. (2016). New TRB Publication: The Highway Capacity Manual, 6th Edition:

A Guide for Multimodal Mobility Analysis. Transportation Research Board, 306, 16–

21.

Elefteriadou, L., Dowling, R. G., & Ryus, P. (2015). Exploring Multimodal Analysis in the

Highway Capacity Manual 2010. ITE Journal, 85(2).

https://trid.trb.org/View/1356820

Elefteriadou, L., Srinivasan, S., Steiner, R. L., Tice, P. C., Lim, K., University of Florida, G.,

& Florida Department of Transportation. (2012). Expanded Transportation

Performance Measures to Supplement Level of Service (LOS) for Growth

Management and Transportation Impact Analysis (01454384; p. 203p).

http://ftp.fdot.gov/file/d/FTP/FDOT%20LTS/CO/research/Completed_Proj/Summary

_PL/FDOT_BDK77_977-14_rpt.pdf

111
FDOT. (2016). Manual on Uniform Minimum Standards for Design, Construction and

Maintenance for Streets and Highways. Florida Department of Transportation.

https://fdotwww.blob.core.windows.net/sitefinity/docs/default-

source/roadway/floridagreenbook/2016floridagreenbookfinal-

982972170.pdf?sfvrsn=946ed802_2

FDOT. (2017). Transit Facilities Guidelines. Florida Department of Transportation.

https://fdotwww.blob.core.windows.net/sitefinity/docs/default-

source/transit/documents/transitfacilityguidelines-8-4-2017.pdf?sfvrsn=6d929e38_2

FDOT. (2020a). FDOT Design Manual; Development and Processes. Florida Department of

Transportation. https://www.fdot.gov/roadway/fdm/2020-fdot-design-manual

FDOT. (2020b). FDOT Context Classification Guide. Florida Department of Transportation.

https://fdotwww.blob.core.windows.net/sitefinity/docs/default-

source/roadway/completestreets/files/fdot-context-

classification.pdf?sfvrsn=12be90da_2

FDOT & Smart Growth America. (2015). Complete Streets Implementation Plan; M2D2:

Multimodal Development and Delivery.

https://fdotwww.blob.core.windows.net/sitefinity/docs/default-

source/roadway/completestreets/files/final-csi-implementation-

plan.pdf?sfvrsn=96979438_2

Grant, M., Rue, H., Trainor, S., Bauer, J., Parks, J., Raulerson, M., Rooney, K., & Suter, S.

(2012). The Role of Transportation Systems Management & Operations in Supporting

Livability and Sustainability: A Primer. United States. Federal Highway

112
 Administration. https://ops.fhwa.dot.gov/publications/

fhwahop12004/fhwahop12004.pdf

Guttenplan, M., Landis, B. W., Crider, L. B., & McLeod, D. S. (2001). Multimodal Level-of-

Service Analysis at Planning Level. Transportation Research Record: Journal of the

Transportation Research Board, 1776, 151–158.

Harvey, J. T., Saboori, A., Butt, A. A., Kendall, A., Ostovar, M., Hernandez, J., Haynes, B.,

National Center for Sustainable Transportation, University of California, D., JCH

Research, California Department of Transportation, & Office of the Assistant

Secretary for Research and Technology. (2018). Framework for Life Cycle

Assessment of Complete Streets Projects (01690028; p. 200p).

https://escholarship.org/uc/item/0vw335dp

Heller, D. S. (2014). Performance Measurement for Transportation Infrastructure: The

Paradigm for Transportation Planning in the 21st Century. 669–681.

https://doi.org/10.1061/9780784413586.064

INRIX. (2019). INRIX: Shared Bikes and Scooters Could Replace Nearly 50 Percent of

Downtown Vehicle Trips. Inrix. https://inrix.com/press-releases/micromobility-study-

us-2019/

Janssen, C., Barbour, W., Hafkenschiel, E., Abkowitz, M., Philip, C., & Work, D. B. (2020).

City-to-city and temporal assessment of peer city scooter policy. Transportation

Research Record, 2674(7), 219–232.

Kahle, D., & Wickham, H. (2013). ggmap: Spatial Visualization with ggplot2. The R Journal,

5(1), 144–161.

113
Kanafani, A., Wang, R., University of California Transportation Center (UCTC), & Research

and Innovative Technology Administration. (2010). Measuring Multimodal Transport

Level of Service. In UCTC Research Paper (01330098).

http://www.uctc.net/research/papers/UCTC-FR-2010-26.pdf

Kingsbury, K., Lowry, M., & Dixon, M. (2011). What makes a “Complete Street” complete?

Transportation Research Record: Journal of the Transportation Research Board,

2245, 103–110. https://doi.org/10.3141/2245-13

Kittelson & Associates, Inc. (2020). Context Classification Framework For Bus Transit.

Florida Department of Transportation.

https://fdotwww.blob.core.windows.net/sitefinity/docs/default-

source/transit/documents/context-classification-framework-for-bus-transit-

final.pdf?sfvrsn=a53e4325_2

Kittelson & Associates, Inc., Texas Transportation Institute, & Transport Consulting Limited.

(1999). Transit Capacity and Quality of Service Manual. Transit Cooperative

Research Program, Transportation Research Board of the National Research Council.

Lasley, P. (2016). Developing Multimodal Performance Measures: Understanding What They

Are and How to Develop Them. Transportation Research Board Annual Meeting,

12p. https://trid.trb.org/view/1393472

Lee, C., & Miller, J. S. (2017). A probability-based indicator for measuring the degree of

multimodality in transportation investments. Transportation Research Part A: Policy

and Practice, 103, pp 377-390.

114
Lekach, S. (n.d.). New app shows you all the e-scooters nearby in one handy map. Mashable.

Retrieved May 10, 2021, from https://mashable.com/article/scooter-map-app-e-

scooter-lime-bird/

Lovas, D., Nabors, D., Goughnour, E., & Rabito, L. (2015). Massachusetts Department of

Transportation Complete Streets Pedestrian and Bicycle Level of Service Study.

Transportation Research Board Annual Meeting, Article 15–4803.

https://trid.trb.org/View/1338857

MassDOT. (2015). Separated Bike Lane Planning and Design Guide. Massachusetts

Department of Transportation. https://www.mass.gov/lists/separated-bike-lane-

planning-design-guide

MassDOT. (2020). Guidelines for the Planning and Design of Roundabouts. Massachusetts

Department of Transportation. https://www.mass.gov/doc/massdot-guidelines-for-the-

planning-and-design-of-roundabouts/download

MassHighway. (2006). Project Development and Design Guide. Massachusetts Department

of Transportation.

MnDOT. (n.d.). Road Design Manual. Minnesota Department of Transportation.

https://edocs-

public.dot.state.mn.us/edocs_public/DMResultSet/download?docId=12772885

MnDOT. (2017). Statewide Multimodal Transportation Plan: 2017 to 2036. Minnesota

Department of Transportation. https://minnesotago.org/final-plans/smtp-final-plan

115
MnDOT. (2018). Performane-Based Practical Design (PBPD) Process and Design

Guidance. Minnesota Department of Transportation. https://edocs-

public.dot.state.mn.us/edocs_public/DMResultSet/download?docId=2156389

MnDOT. (2020). Bicycle Facility Design Manual. Minnesota Department of Transportation.

https://edocs-

public.dot.state.mn.us/edocs_public/DMResultSet/download?docId=6976095

Neuman, T. R., Coakley, R. C., Panguluri, S., Harwood, D. W., National Cooperative

Highway Research Program, Transportation Research Board, & National Academies

of Sciences, Engineering, and Medicine. (2016). A Performance-Based Highway

Geometric Design Process (p. 24626). Transportation Research Board.

https://doi.org/10.17226/24626

Ni, Y., Ling, Z., & Li, K. (2013). A New Evaluation Method Combining Efficiency and

Safety: Multimodal Comprehensive Level of Service of Signalized Intersections.

1447–1455. https://doi.org/10.1061/9780784413036.193

ODOT. (2012). Highway Design Manual. Oregon Department of Transportation.

https://www.oregon.gov/odot/Engineering/Documents_RoadwayEng/HDM_00-Full-

Report.pdf

ODOT. (2017). Project Delivery Guide. Oregon Department of Transportation.

https://www.oregon.gov/odot/ProjectDel/Documents/SBS-Guidebook.pdf

ODOT. (2020a). Blueprint for Urban Design (BUD)—Vol 1 (Vol. 1). Oregon Department of

Transportation.

https://www.oregon.gov/ODOT/Engineering/Documents_RoadwayEng/Blueprint-for-

Urban-Design_v1.pdf

116
ODOT. (2020b). Blueprint for Urban Design (BUD)—Vol 2 (Vol. 2). Oregon Department of

Transportation.

https://www.oregon.gov/ODOT/Engineering/Documents_RoadwayEng/Blueprint-for-

Urban-Design_v2.pdf

Pathak, C., Arabkhedri, B., & MacKenzie, D. (2020). Misparked Repo.

https://misparkrepo.com/

Pathak, C., Arabkhedri, B., & MacKenzie, D. (2021). Crowd-sourcing Micro-mobility

Parking Violation Reporting – User Interface Design Motivation and Analytical

Opportunities from Data Collected. Transportation Research Board Annual Meeting.

Phillips, R., Karachepone, J., Landis, B., University of Florida, G., & Florida Department of

Transportation. (2001). Multi-Modal Quality of Service Project (00808643; p. 117 p.).

https://rosap.ntl.bts.gov/view/dot/15408

Pontis, V. (n.d.). Scooter Map—All the Scooters on One Map. Retrieved May 11, 2021, from

https://scootermap.com/

R Core Team. (2021). R: A language and environment for statistical computing [Manual].

https://www.R-project.org

Ray, B. L., Ferguson, E. M., Knudsen, J. K., Porter, R. J., Mason, J., National Cooperative

Highway Research Program, Transportation Research Board, & National Academies

of Sciences, Engineering, and Medicine. (2014). Performance-Based Analysis of

Geometric Design of Highways and Streets (p. 22285). Transportation Research

Board. https://doi.org/10.17226/22285

117
Sanders, R., Macdonald, E., & Anderson, A. (2010). Performance Measures for Complete,

Green Streets: A Proposal for Urban Arterials in California. Transportation Research

Board Annual Meeting, 17p. https://escholarship.org/uc/item/54h8k27x.

https://trid.trb.org/view/910237

Seskin, S., Kite, H., Searfoss, L., AARP, & Smart Growth America. (2015). Evaluating

Complete Streets Projects: A Guide for Practitioners (01567388; p. 49p).

http://www.smartgrowthamerica.org/documents/evaluating-complete-streets-

projects.pdf

Sinclair, C., Schaefer, R., & others. (2019). Multimodal System Performance Measures

Research and Application: Innovation and Research Plan. United States. Federal

Highway Administration. https://rosap.ntl.bts.gov/view/dot/43579

Tasic, I., & Bozic, C. (2017). Toward True Multimodal Transportation Accessibility: Data,

Measures, and Methods. Transportation Research Board Annual Meeting.

https://trid.trb.org/view/1438514

Tennekes, M. (2018). tmap: Thematic maps in R. Journal of Statistical Software, 84(6), 1–39.

https://doi.org/10.18637/jss.v084.i06

Transportation Statistics Office, Florida Department of Transportation. (2015). Florida

Multimodal Mobility Performance Measures Source Book.

Tufte, E. R. (1985). The visual display of quantitative information. The Journal for

Healthcare Quality (JHQ), 7(3), 15.

Venables, W. N., & Ripley, B. D. (2002). Modern applied statistics with s (4th ed.). Springer.

https://www.stats.ox.ac.uk/pub/MASS4/

118
Walker, K. (2020). tidycensus: Load US census boundary and attribute data as “tidyverse”

and ’sf’-Ready data frames. https://CRAN.R-project.org/package=tidycensus

Walker, K. E. (2016). tigris: An R package to access and work with geographic data from the

US Census Bureau. https://cran.r-project.org/web/packages/tigris/index.html

Wang, L., Liu, J., Portland State University, Oregon Department of Transportation, & Federal

Highway Administration. (2016). Multi-Modal Performance Measures in Oregon:

Developing a Transportation Cost Index Based Upon Multi-Modal Network and Land

Use Information (01593722; p. 132p).

http://www.oregon.gov/ODOT/Programs/ResearchDocuments/SPR760_TCIFinalRep

ort.pdf

Washington State Department of Transportation. (2014). WSDOT’s Handbook for Corridor

Capacity Evaluation (01544655; p. 48p).

http://wsdot.wa.gov/publications/fulltext/graynotebook/CCR14_methodology.pdf

Zuniga-Garcia, N., Ross, H. W., & Machemehl, R. B. (2018). Multimodal Level of Service

Methodologies: Evaluation of the Multimodal Performance of Arterial Corridors.

Transportation Research Record: Journal of the Transportation Research Board,

2672(15), 142–154. https://doi.org/10.1177/0361198118776112

119
Appendices

120
Appendix A

Lists of micromobility infraction rules for eight different cities are brought in Tables 16 to 23.

The synthesis of all the infraction rules can be seen together in Table 13. The severities were

assigned by the researchers of the study. These scores are to some extent subjective based on

how much we thought a certain issue had priority. The infraction rules themselves, however,

are entirely based on permits studies online.

Table 16: List of micromobility infraction rules for Seattle, WA

121
Table 17: List of micromobility infraction rules for Portland, OR

122
Table 18: List of micromobility infraction rules for Redmond, WA

Table 19: List of micromobility infraction rules for Spokane, WA

123
 Table 20: List of micromobility infraction rules for Boise, ID

Table 21: List of micromobility infraction rules for San Francisco, CA

124
Table 22: List of micromobility infraction rules for Los Angeles, CA

Table 23: List of micromobility infraction rules for Washington D.C.

125
 ProQuest Number: 28643109

INFORMATION TO ALL USERS

The quality and completeness of this reproduction is dependent on the quality
and completeness of the copy made available to ProQuest.

Distributed by ProQuest LLC ( 2021 ).

Copyright of the Dissertation is held by the Author unless otherwise noted.

This work may be used in accordance with the terms of the Creative Commons license
or other rights statement, as indicated in the copyright statement or in the metadata
associated with this work. Unless otherwise specified in the copyright statement
or the metadata, all rights are reserved by the copyright holder.

This work is protected against unauthorized copying under Title 17,

United States Code and other applicable copyright laws.

Microform Edition where available © ProQuest LLC. No reproduction or digitization

of the Microform Edition is authorized without permission of ProQuest LLC.

ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346 USA