Transportation Research Part D 103 (2022) 103173

Contents lists available at ScienceDirect

Transportation Research Part D

journal homepage: www.elsevier.com/locate/trd

Congestion and environmental impacts of short car trip

replacement with micromobility modes
Zhufeng Fan a, *, Corey D. Harper a, b
a
Department of Civil and Environmental Engineering, Porter Hall, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA
b
Heinz School of Information Systems and Public Policy, Hamburg Hall, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213,
USA

A R T I C L E I N F O A B S T R A C T

Keywords: Transportation is a basic social need, but most trips are done by private vehicles, which is not
Micromobility environmentally sustainable with growing urban populations. Micromobility (e.g., shared bikes)
Travel Demand Model represents a significant opportunity to replace short private vehicles trips (0–3 miles) and reduce
Congestion
transportation sector emissions. This paper uses Seattle as a case study and estimates that up to
Emissions
Energy Use
18% of short car trips could be replaced by micromobility. A static traffic assignment model is
developed to simulate and compare the results of peak hour traffic under a base case scenario
(2014 traffic conditions) to scenarios where a portion of short car trips are substituted by
micromobility. Results indicate that micromobility could reduce congestion on heavily congested
corridors and wide-scale bike lane deployment can maximize congestion benefits, but the impacts
to energy use and emissions are disproportionately low and other measures (e.g., vehicle elec­
trification) are needed to meet climate change emissions targets.

1. Introduction

Transportation is a basic social and economic need, but most trips are done by private vehicle, which is not economically or
environmentally sustainable with growing urban populations (United Nations Department of Economic and Social Affairs, 2014).
From 2001 to 2019, total light-duty vehicle miles traveled (VMT) on urban roadways in the United States has increased by about 26%
(from 1.7 trillion miles to 2.3 trillion miles) (USDOT Federal Highway Administration, 2021), while the total urban population in the
United States has increased by 20%, over the same period (The World Bank, 2021). As a result, more commuters are spending increased
amounts of time in congestion, which has associated costs such as wasted time, money, fuel, and emissions. For example, in 2019,
Americans, on average, lost 99 h a year due to congestion, costing them nearly $88 billion, an average of $1,377 per driver (Schrank
et al., 2019). Cities have tried to solve congestion issues using several traffic management strategies such as traffic light optimization
(StoIlova and StoIlov, 1998), lane expansion (Cervero, 2003), and bus service improvements (Li et al., 2021), but many of these so­
lutions tend to be costly and even with these measures implemented, the issue of congestion remains unresolved in many urban areas
(Blethen, 2019).
According to the United States Department of Transportation (USDOT), the average U.S. household produces about 9.5 trips a day
(USDOT Federal Highway Administration, 2021). About half of these trips are within three miles, but fewer than 2 percent of those
trips are made by bicycles. Private vehicles such as cars, pick-up trucks, and SUVs account for 50 percent of short distance trips (i.e.,

* Corresponding author.
E-mail addresses: zhufengf@andrew.cmu.edu (Z. Fan), cdharper@andrew.cmu.edu (C.D. Harper).

https://doi.org/10.1016/j.trd.2022.103173

Available online 2 February 2022

1361-9209/© 2022 Elsevier Ltd. All rights reserved.
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

trips 3 miles or less) in most US metropolitan areas (INRIX, 2019). Micromobility (defined as docked or dockless shared bikes, e-bikes,
scooters, e-scooters, skateboards, etc.) (Martin, 2021) represents a significant opportunity to replace short POV trips and reduce
transportation sector greenhouse gas (GHG) emissions (Shaheen et al., 2010). In the past decades, cycling has experienced a “re­
naissance” (Pucher et al., 2011) and the concept of shared micromobility has spread all over the world (Shaheen et al., 2010).
As of 2019, there were over 400 bike-sharing systems in operation, including 103 docked systems and 71 dockless systems, and 239
e-scooter systems in 192 cities all over the US (US Bureau of Transportation Statistics, 2020). In 2019, people in the United States had
136 million trips using micromobility modes, a 60% increase when compared to the previous year (National Association of City
Transportation Officials, 2020). Shared micromobility, while not climate neutral, does use less energy and emit less GHG emissions per
person-kilometer over their life cycle when compared to private cars. In addition to the environmental and societal benefits (from
reduced emissions) of bike share programs, there is little difference in speed between cars and bikes in urban arterials under current
peak hour congestion. The average automobile speed of urban areas in 978 cities in 7 continents is about 7mph calculated from INRIX
2019 Global Traffic Scorecard (INRIX, 2020), whereas Mobike users’ average cycling speed is 3.2–6 mph (Mobike, 2018a). Considering
far less road space occupancy than cars, and the ability to meet similar purposes for short trips for modern cities, micromobility is seen
as a viable and low-cost alternative to replacing many short car trips in urban areas.
This paper develops a methodology for estimating the upper bound number of short-distance private vehicle trips that could be
replaced by micromobility modes and applies this method to a case study in Seattle, WA to estimate the resulting environmental and
congestion benefits. These estimates are based on a static traffic assignment model, which is used to assess how peak hour traffic
conditions could change in scenarios where micromobility has replaced a portion of short car trips.

2. Literature review

2.1. Statistical methods

Statistical methods are widely to assess the effects of micromobility on traffic congestion, emissions, and energy use. Wang and
Zhou (2017) used roadway inventory data from the Highway Performance Monitoring System along with weather and climate data to
study the influence of shared bikes on congestion in 96 urban areas (Wang and Zhou, 2017). Wang and Zhou found that a 1% increase
in population growth in cities with shared bikes would result in 0.026% less congestion when compared to cities without shared bikes.
Hamilton and Wichman (2018) investigated the effects of shared bike infrastructure on traffic congestion in Washington DC using
panel data model regression (Hamilton and Wichman, 2018). Hamilton and Wichman estimated that bike-sharing can reduce
congestion upwards of 4% within a neighborhood and concluded that congestion alleviation is concentrated more in highly congested
areas. Brand et al. (2021) applied a linear regression model to longitudinal travel data from seven European cities and estimated that
an average person shifting modes from car to bike decreased emissions by 3.2 kg CO2/day (Brand et al., 2021). Fan and Zheng (2020)
evaluated the effects of micromobility on congestion using a difference in difference method to compare average rush hour congestion
before and after the implementation of shared bikes in Beijing, China (Fan and Zheng, 2020). Fan and Zheng estimated that dockless
bike sharing could facilitate the usage of the subway system and mitigate traffic congestion surrounding subway stations by about 4%.
Researchers have also combined statistical analysis with other methods to estimate the impacts of micromobility replacement on
congestion and the environment. For example, Zahabi et al. (2016) adapted a binary logistic regression model and simultaneous
equation modeling approach to the present commuting choice and household location (Zahabi et al., 2016). Zahabi et al. found that an
increase in bike infrastructure accessibility could facilitate ridership and that the addition of new bicycle facilities could result in a 2%
decrease in GHG emissions. Li et al (2021) utilized a mixed logit model to estimate the mode substituted by dockless bike share and
assessed the resulting GHG emission reduction using life cycle analysis (LCA) (Li et al., 2021). Their analysis is based on trip level data
from Mobike that covers 14 consecutive days. Li et al., estimated that each dockless bike share trip saves about 81 g CO2 equivalent
GHG emissions on average and 117 kilotons annually in the city of Shanghai.

2.2. Other methods (e.g., surveys, scenario analysis, and life cycle assessments)

Researchers have also utilized other methods such as surveys and scenario analysis to assess the effects of micromobility on
congestion and/or environmental sustainability. For example, Hiselius and Svensson (2017) sent an online survey to e-bike users in
Sweden and found that changes in transport mode (i.e., switching from car to e-bike) reduced CO2 emissions by 8.2 kg/week, on
average (Winslott Hiselius and Svensson, 2017). Lovelace et al. (2011) provided predictive energy savings for a set of future scenarios
with increased cycling rates in Sheffield between 2010 and 2020, based on empirical data prior to 2009. The results indicate that
replacing all possible car trips by bikes can save 48.3 trillion Joules energy (90.3 million Joules per capita) (Lovelace et al., 2011). Kou
et al. (2020) proposed a bike share emission reduction estimation model to quantify the environmental benefits of bike share trips and
compare GHG emission reductions from bike share systems in eight US cities (Kou et al., 2020). Kou et al. (2020) reveal annual
emissions reduction from the eight cities range anywhere between 41 tons (Seattle) to 5,417 tons (New York City) CO2 equivalent, and
283 to 581 g CO2 equivalent per trip. Cherry et al. (2009) quantified the life-cycle emissions of several transportation modes in China
and estimated that electric two-wheelers produce several times fewer emissions than cars and motorcycles per passenger kilometer
(Cherry et al., 2009). Ji et al. (2012) demonstrated that replacing conventional automobiles by e-bikes can reduce environmental
health impacts by several times (gasoline cars 2×, diesel cars 10×, and diesel buses 5 × ) in terms of PM2.5 (Ji et al., 2012). Mason et al.
(2015) depicted the future world of cycling by creating a high shift cycling scenario in global cities, showing that cycling can save 24
trillion dollars from 2015 to 2050, and reduce 11 percent of CO2 emissions and energy use from entire urban transportation in 2050

2
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

(Mason et al., 2015). Their results also indicate that bicycles and e-bikes will share 14% of miles travel in urban areas in global cities,
on average. Accordingly, in alignment with this share rate estimation, McQueen et al. developed a mode replacement model in
Portland, Oregon using a self-reported survey data of electric bikes to automobiles (McQueen et al., 2020). McQueen et al. estimated
that increasing e-bike mode share by person miles traveledtraveled (PMT) to 15%, could result in a 12% reduction in Portland’s
passenger transportation emissions. Rojas-Rueda et al. (2012) estimated that a 40% mode shift from car to micromobility could reduce
CO2 emissions by 200 kilotons per year in the Barcelona Metropolitan area (Rojas-Rueda et al., 2012). Zhang and Mi (2018) utilized the
Mobike dataset in the city of Shanghai, China to estimate the environmental benefits of shared bikes (Zhang and Mi, 2018). Zhang and
Mi estimate trip distances between origination and destinations of bikes then compute petrol consumption as well as CO2 and NOx
emissions by vehicles if the bike trip distances are greater than 1 km (0.62 miles). Their results suggest that bike sharing in Shanghai
saved 25,240 tons of CO2, 64 tons of NOx, and 8,400 tons of petrol in 2016.
It is also indicated that micromobilty has life-cycle emissions from manufacturing, transporting, to repositioning for shared bikes.
Hollingsworth et al. (2019) estimated environmental burdens associated with dockless e-scooters using life cycle assessment (Hol­
lingsworth et al., 2019). Hollingsworth et al. found that replacing personal vehicle travel with e-scooter usage would result in a net
reduction in environmental impacts. Fishman et al. (2014) investigated how many car trips can be replaced by bike share by examining
survey and trip data from several bike share programs, and the resulting net change of vehicle trips from the mode replacement
(Fishman et al., 2014). Fishman et al. observed that every kilometer car trip replaced by shared bikes may cause 0.2 to 2 km extra car
trips because of shared bikes rebalancing work. De Bortoli and Christoforou (2020) applied consequential life cycle assessment to
estimate the carbon footprint of modal shifts to dockless e-scooters in Paris, France (de Bortoli and Christoforou, 2020). De Bortoli and
Christoforou found that dockless e-scooters may produce more life-cycle emissions per person-kilometer than the Paris metro transit
system, but less per person-kilometer emissions than private cars and buses. Reck et al. (2021) built a mode choice model and esti­
mated the possible mode replacement rate of private and shared e-bike and e-scooters over walk, transit, car, and bike, and the
consequential life cycle net emission changes (Reck et al., 2021). Reck et al. found that personal micromobility emits less CO2 than the
transport modes they replace, while shared micromobility emits more CO2 than the transport modes they replace.
The aforementioned studies have demonstrated the great potential for micromobility to alleviate congestion and reduce emissions

Fig. 1. Case Study Area and Seattle Road Network.

3
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

and energy usage in urban areas using several different methods. However, there is still not a general method to estimate to what extent
such substitution could happen, as well a detailed traffic model to reflect the resulting congestion and environmental benefits. This
paper makes a contribution to the literature by developing a method to estimate the upper bound number of short private vehicle trips
with micromobility using household travel survey and weather data and using a static traffic assignment to assess changes in traffic
conditions, energy use, and emissions after mode replacement. This framework can be used by public agencies (e.g., Metropolitan
Planning Organizations) to incorporate micromobility modes into the long-range planning of the transportation system.

3. Case study

This paper focuses on assessing traffic impacts in Seattle, WA, in large part due to the fact that is an urban area with one of the
greatest potentials to replace short car trips with micromobility modes, as currently, short car trips account for 48% of total peak hour
trips (Reed, 2019). As one of the most congested cities in the US (Merten, 2020) and has suitable weather for cycling, the city of Seattle
is also promoting the widespread use of micromobility (Seattle DOT, 2021). In order to capture trips that start and/or end the city of
Seattle, we consider travel demand and road network data across the Puget Sound region, which includes four counties (King; Pierce;
Snohomish; and Kitsap) (Fig. 1).
In Seattle, there are a total of 864 freeway/expressway road links and 8,213 urban arterial links. The major freeway/expressways
that provide access to and pass through the city of Seattle are Interstate-5, Interstate-90, State Route-99, and State Route-520.
Additionally, car ferries are also important in connecting Seattle and its west islands across the Puget Sound. Car ferry routes
include State Route-304 and State Route-305. The network in Seattle we used for traffic simulation has car-only accessible roads, bike
lanes, and urban arterials where bikes have the same right as vehicles since the congestion impacts stem from traffic performance and
environmental implications is based on the difference of vehicle miles traveled after mode substitution.

4. Data

4.1. Data for estimating upper bound potential for micromobility to replace short car trips

4.1.1. 2014 Puget Sound Regional Council household travel survey data
The primary data source used to estimate the upper bound number of short-car trips that could be replaced with micromobility
modes is the Puget Sound Regional Council (PSRC) 2014 Spring Household Survey (Puget Sound Regional Council, 2016). PSRC
periodically releases information on the travel and transportation characteristics of the Puget Sound Region by conducting a repre­
sentative regional survey, to assist policymakers and transportation planners in quantifying travel behavior and analyzing changes in
travel characteristics over time.
The 2014 PSRC Household Survey Data were collected via questionnaires, filled by an adult member in households that partici­
pated in the survey. The questionnaire was designed for participants to report three general types of information: 1) household,
vehicle, and person-level demographics; 2) the 24-hour travel dairy on the assigned travel date; 3) travel behaviors and attitudes. This
dataset is a representative sample with a sample size of about 6,000 households for the PSRC region (King, Kitsap, Pierce, and Sno­
homish). To reproduce typical weekday travel patterns, survey participants were assigned to a travel day between April 8 to June 12
for all seven days of the week including holidays. Respondents were asked to report all trips made within the household on their
assigned travel day. For each trip entry, respondents were asked to report trip mode, trip purpose, distance, and the time of day the trip
took place. The origin and destination locations are accurate to the census tract group level. Zipcode, Public Use Microdata Areas
(PUMA), city name, and county information are also provided in the dataset.
The 2014 PSRC Household Travel Survey attempts to represent the travel characteristics of the Puget Sound Region on a regional
level. A weighting factor is provided for each person, household, trip, and vehicle included in the datasets. This weighting factor is the
computed inference factor, which is intended to represent the total population from which the sample was drawn. Any analysis done
using the survey was conducted using the provided weighted factors. Detailed information regarding trip or person-level data can be
found in PSRC 2014 Household Survey Documents (Puget Sound Regional Council, 2016).

4.1.2. Darksky weather data

In addition to travel survey data, weather is also an important factor when estimating the upper bound micromobility replacement
potential, since a short car trip can only be replaced in good weather conditions. In this research, we scraped weather data from the
Darksky API (Darksky, 2021), which records hourly historical weather information in Seattle covering the whole travel survey data
collection time period (Spring 2014). Weather features of concern are precipitation and wind speed. More information on how this
weather data was utilized can be found in the Methods section of the paper.

4.2. Data for static traffic assignment model

4.2.1. Puget Sound Regional Council travel demand model

In this analysis, we leverage the PSRC’s regional planning model to develop our traffic assignment model, which contains an urban
road network and travel demands containing origination-destination trips information across all the four counties mentioned in Fig. 1.
The PSRC road network is the basis of traffic simulations for regional planning in the Puget Sound Region. The PSRC road network is a
simplified version of the true road network. While it contains all the primary and secondary roads, it aggregates tertiary and local roads

4
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

into simplified road segment representations to make traffic simulations computationally efficient. The PSRC road network does not
represent toll stations, carpool or express lanes, traffic signals, and yield/stop signs, however, HOV links are indicated, and intersection
delay caused by traffic lights and signs are included as a delay term in volume delay functions applied on all links. The network consists
of 55,545 directed road links, 25,596 intersection nodes, 155,296 turns, 14,946 connectors, and 3,700 traffic analysis zones (TAZs).
For each link, information on the posted speed limit, allowable travel modes, hourly lane capacity, link tolls, and volume delay
functions are provided.
Demand data, which contains information on the number of trips made by private vehicles, trucks, bikes, light rail, and walking for
each unique origin–destination pair during PM peak hours, was also provided by PSRC. This data was generated using activity-based
models for the year 2014 to represent trips in a typical workday. Private vehicle demands are provided in terms of Single Occupancy
Vehicles (SOV) trips and High Occupancy Vehicles (HOV2 and HOV3) trips for each origin–destination pair. Trucks demands are
provided in terms of medium trucks and heavy trucks, both as Passenger-Car-Equivalent (PCE) trips. During peak hour travel times,
over 70% of trips in the Puget Sound Region are done by private vehicles, 20% by walking, and the rest by other modes (e.g., biking and
transit).

4.2.2. Model calibration and validation data

Two data sources are used for model calibration and validation:1) Washington State Department of Transportation (WSDOT) 2014
hourly freeway traffic count data and 2) Seattle Department of Transportation 2014 Traffic Flow Counts (Seattle GeoData Portal,
2019). The WSDOT hourly freeway traffic count dataset contains data on the total number of vehicles traveling on highways in the
Puget Sound Region during peak hour travel times. The 2014 Traffic Flow Counts dataset contains traffic counts data for main arterials
in Seattle. These datasets are used to calibrate and validate our base model against observed traffic counts. Our final calibration and
validation dataset consisted of 392 total links, including 149 urban arterials and 243 freeway observation counts.

5. Methods

In order to estimate the congestion, energy use, and emissions impacts of replacing short car trips with micromobility modes, we
follow the overall procedure shown in Fig. 2. Our proposed method has two main components: 1) data analysis and 2) traffic simu­
lation. The former aims at estimating the upper bound short car trip replacement rate taking into account trip (e.g., trip purpose) and
person data (e.g., age), and weather conditions (e.g., precipitation). We first gather Seattle household survey data and weather data,

Fig. 2. Overall Research Method.

5
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

identify short trips done by private vehicles, and estimate the upper bound potential for micromobility to replace short car trips.
Demand scenarios, from 0% short car trip replacement to the upper bound penetration rate are incorporated into the static traffic
assignment model and compared to base case conditions. Finally, we simulate updated traffic conditions based on updated demand
scenarios, and discuss the resulting congestion, emissions, and energy use benefits.

5.1. Estimating upper bound of micromobility replacement

To estimate the upper bound number of short car trips that can be replaced by micromobility we follow the procedure shown in
Fig. 3. As mentioned earlier, the main sources of data to conduct the upper bound estimation analysis are the 2014 PSRC Household
Travel Survey and DarkSky weather data. For this analysis, we only consider short trips made by private vehicles during weekday PM
peak hours (3 pm-6 pm). Trips that took place on a federal holiday (e.g., Memorial Day) were not considered. The importance of
estimating this upper bound stems from the fact that all short trips cannot be replaced due to trip, person, and weather limitations that
prevent micromobility from being a viable mode for certain trips.
To identify short trips that are possible to be replaced by micromobility, we consider the entire tour that includes that short trip. A
tour includes all trips from leaving home to returning home, so a person may have several tours in a day. Several requirements have
been set up for the tour. First, all trips in a tour must be a short trip (i.e., less than 3 miles), because the whole tour needs to be replaced

Fig. 3. Methodology for Estimating the Upper Bound Potential to Replace Short Car.

6
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

by micromobility mode, as a person who leaves home with his/her personal vehicle must be able to return home using the same mode
of transport. Second, the person on the trip is older than 16 years old. This is because most shared micromobility operators require
users to be 16 years older to ride a bike, such as Lime (Lime, 2021) and Mobike (Mobike, 2018b). Because vendors do not have an upper
cut off age limit and research shows that cycling could improve the health and mobility of people over the age of 65 (Johnson and Rose,
2015) (Van Cauwenberg et al., 2019), we do not establish an upper cut off limit for age. Third, we consider wind effects on cycling, as
air drag is a major resistance that could consume power generated by cyclists, and such effect is highly dependent on wind speed
(Saneinejad et al., 2012) (Isvan, 2015) (Belloli et al., 2016). There is no authoritative resource indicating an appropriate cycling wind
speed, but empirically, cycling in over 20 mph wind is not guaranteed to be safe and controllable (Matheny, 2019) (Carmichael, 2020)
(Sears et al., 2012). The highest wind speed during the trip time in March-June 2014 is 10.42 mph which is Gentle Breeze in Beaufort
Wind Scale (American Meteorological Society, 2015) and suitable for cycling, therefore wind would not adversely impact peak hour
micromobility trips in Seattle. If this framework were applied to other cities where strong wind frequently occurs, wind speed could
play a major role in the number of short private car trips that could be replaced by micromobility. Fourth, it is assumed that
micromobility will not be an attractive travel mode during most inclement weather conditions (Nankervis, 1999) (Noland and Ishaque,
2006) (Gebhart and Noland, 2014) (Noland, 2021), therefore if a short car trip occurs in moderate rain weather conditions or greater
we assume that none of the trips in the tour are eligible for micromobility replacement. Because this is an upper bound estimate and
people do still bike in inclement weather conditions (Schmiedeskamp and Zhao, 2016) (Zhao et al., 2019), although at lower fre­
quencies, we consider short car trips that occurred in drizzling and light rain weather conditions as eligible for micromobility
replacement. Given that Seattle is a humid city and light rain is the dominant rainfall intensity (Trenberth and Zhang, 2018) (Church,
1974), for upper bound estimation, the possibility for cycling in light rain should also be included. In other words, we consider the fact
that cyclists can be prepared for biking in common wet weather conditions. Lastly, trip purposes are also considered as part of this
framework. In the 2014 PSRC Household Travel Survey, trip purposes include: a) Go home, b) Go to workspace, c) Go grocery

Fig. 4. Micromobility Penetration Rate Distribution by Public Use Microdata Areas Across Seattle, WA.

7
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

shopping, d) Go other shopping (e.g., mall, pet store, etc.), e) Go to school/daycare, f) Go to medical appointment, g) Conduct personal
business, h) Drop off/pick up someone, i) Go exercise, j) Go to restaurant, k) Attend social event, l) Attend recreational event, m) Go to
religious/community/volunteer activity, n) Transfer to another mode of transportation; and o) Other (Puget Sound Regional Council,
2016). Among these trips, we assume that picking up / dropping off a person cannot be replaced, since micromobilty is a single-
occupancy mode of travel and therefore, a single bike/scooter cannot be used to safely transport multiple passengers at once. We
also assume that shopping trips cannot be replaced by micromobility modes, as transporting goods with bike or scooter can be
cumbersome. Therefore, if there is any shopping or picking up / dropping off trip purpose in a tour, we assume that none of the trips in
the tour can be replaced.

Trips with micromobility modes

Because the upper bound potential for micromobility to replace short car trips may vary in different parts of the city and the
penetration rate may not be uniformly distributed, we estimate an upper bound for each Public Use Microdata Area (PUMA) (US
Census Bureau, 2021). We select origin PUMA as our estimate unit because 1) neither zipcodes nor census block groups are able to
comprise enough trip samples to generate a reasonable estimate, 2) PUMA boundaries can appropriately split the city of Seattle and
critical parts such as downtown, central business district, university place, and rural areas are located in different PUMAs, 3) each
PUMA has a similar amount of population (no fewer than 100,000 people), and 4) The boundary of PUMA exactly contains boundaries
of TAZs in our traffic simulation analysis, so we can easily apply the upper bound estimate (i.e., replacement rate) for every PUMA to
TAZ level according to location and modify the simulation demand matrix based on TAZ. The upper bound estimate for each PUMA is
estimated using the 2014 PSRC Travel Survey, where we assess the number of short trips that originated in each PUMA and compare
this number to the number of short trips that could be replaced by micromobility modes. We calculated the upper bound potential on
all PUMAs in the PSRC region, not limited to the city of Seattle which is our analysis focusing area. This is because the upper bound
mode switch potential is finally adopted to the trips originated from every TAZ in the travel demand model, which is necessary to
accurately simulate traffic simulation in the city of Seattle. Fig. 4 demonstrated the micromobility penetration rate distribution by
PUMAs across the city of Seattle. As can be seen from the picture, the central business district tends to have higher portion of trips
viable to be replaced than other parts.
Those trips meeting the outlined criteria constitute approximate 18% of short trips in the PSRC region, in average on all PUMAs. But
not every PUMA has the same potential for replacing short private car trips with micromobility, as the estimate may be affected by
other factors such as demographics and built environment information that are implicitly contained within the survey data for every
PUMA. The estimation means that on average, up to 18% of short private vehicle trips can be substituted by micromobility modes. In
other words, we can also say that the upper bound penetration rate for micromobility systems to replace short car trips is 18% in
Seattle, using the filtering framework that we developed, under 2014 transportation situation.

5.2. Traffic simulation

A traffic simulation model is developed following the steps below: First, we simulate typical workday PM peak hour congestion in
2014 and calibrate the model using observed traffic counts in Seattle. Second, we identify which trips are short trips, and simulate
scenarios where micromobility modes have replaced a proportion of car tips from 0% to the upper bound potential. Finally, we analyze
changes in VMT, as well as the resulting congestion and environmental benefits.

5.2.1. Base case scenario simulation

Traffic simulation is conducted in the four counties that make up the Puget Sound Region, as described earlier in Case Study section.
In this research, a static traffic assignment model is developed using PTV Visum, a traffic planning software. Our first step is to
reproduce current 2014 peak hour traffic conditions, which we will refer to as our base case scenario. Our base case scenario is
developed using the same road network and travel demand data as PSRC’s model and following majority of their network settings.
Private and truck transportation demands are loaded into Visum as five origin–destination matrices (SOV, HOV2, HOV3, medium
trucks, and heavy trucks). Although, bike trips can be converted into a PCE using the Highway Capacity Manual (TRB Highway Ca­
pacity Manual, 2016) or following some proposed methods (Shalini, 2014), bike trips are a very small portion of trips in Seattle when
compared to car trips, so neither PSRC nor us have simulated bikes in the base case scenario. We utilize custom Bureau of Public Roads
volume delay functions provided by PSRC for both urban arterials and highway facilities. To model queues and congestion at in­
tersections, PSRC provided us with arterial delay values, which were incorporated into the custom volume delay functions. Static
user’s equilibrium traffic assignment is conducted to simulate traffic congestion in the Seattle metropolitan.

5.2.2. Base case model calibration and validation

As described in Data section, observed freeway and urban arterial traffic count data is used for base case model calibration and
validation. Observed freeway counts are reported for each hour of the day for each travel direction. Observed urban arterial traffic
counts are reported in terms of AAWDT and converted to peak hour volumes using a PM peak hour volume adjustment (k-factor) and a
directional distribution factor (d-factor). For example, we select a k-factor of 0.065 for the 3–4 PM peak hour according to the temporal
distribution of highway traffic counts. We select a d-factor of 55% and assume that during PM peak hours majority of cars are heading
in the opposite direction of the central business district and towards the highways and main arterials. The origin–destination matrices
were updated using VISUM’s T-Flow Fuzzy demand adjustment method. Model calibration is performed by comparing the observed
link volume to those obtained through simulation.

8
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

The following metrics are used to measure the goodness-of-fit of simulation volume to observation traffic counts, where Y obs is
observed traffic volume counts and Y sim is simulated traffic volume counts from our traffic assignment model, and N is the number of
observation points. Geoffrey E. Havers (GEH) statistic is calculated at each validation data point, where less than 5 is acceptable. We
use the percentage of acceptable GEH statistics to measure the model performance. According to Design Manual for Roads and Bridges,
over 85% of the volumes should have an acceptable GEH statistic. The applied statistics are estimated as follows:

• Mean Relative Percentage Error (MAPE) (Equation (1)):

∑N Ysim − Yobs
i=1 | |
i i
Yobs
MAPE = i
(1)
N

• Relative Root Mean Squared Error (RRMSE) (Equation (2)):

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∑N sim 2
i=1 (Y i − Y obs i ) /(N − 1)
RRMSE = ∑N obs (2)
i=1 Yi /N

• R-square (Equation (3)):

∑N sim 2 ∑N
i=1 (Yi − Yobs
i )
obs
n=1 Yi
R2 = 1 − ∑N 2
, whereY = (3)
sim
− Y) N
i=1 (Yi

• Geoffrey E. Havers (GEH) Statistic (Equation (4)):

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
2(Ysim − Yobs )2
GEH = (4)
Ysim + Yobs
Fig. 5 is a goodness-of-fit visualization during the 3–4 pm peak hour, for expressways/freeways and urban arterials in the Seattle,
after the model calibration process. It is shown that under base case scenario, our model can reproduce the realistic traffic situation at
PM peak hours in a typical workday in the year 2014 for every metric measurement. Therefore, we can assume that scenarios applied to
this model will produce credible results.

5.2.3. Replacing short car trips with micromobility

Mode substitution in traffic simulation is done using the demand matrix, which details the number of trips by different modes for
each TAZ origin–destination pair. Given that a TAZ is a smaller geographic unit than PUMA and there is no overlapping between TAZs

Fig. 5. Calibration of Base Case Model (3-4PM). Note: MAPE = Mean Average Prediction Error, RRMSE = Relative Root Mean Squared Error, and
GEH = Geoffrey E. Havers Statistic.

9
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

and PUMAs, we apply the upper bound potential of each PUMA where the origin TAZ is located to each short trip origin–destination
pair.
To realistically assess how replacing short car trips with micromobility affects congestion, we integrate 2014 bike lane data into the
traffic simulation. Our road network data contains detailed information on the allowable traffic modes of every directional link. For
example, some links allow both bikes and car, some are bike-only links, and others are for vehicular travel only. Given this information,
we developed a methodology, considering three cases, to better assess the congestion effects of replacing short vehicle trips with a
micromobility mode.
Case 1) If there exists a designated or protected bike lane route between an origin–destination pair, we assume the bike trips would
not affect private vehicle traffic stream on vehicular roads, and remove the entire trip from the trip table. Case 2) If there is no
designated bike lane in place to accommodate micromobility traffic between an origin–destination pair, but bikes are allowed to share
the roads with cars (WSDOT, 2021), the replaced short vehicle trips will be converted to micromobility trips in the model, where each
micromobility trip = 0.5 passenger car equivalent, following the Highway Capacity Manual (TRB Highway Capacity Manual, 2016).
Consequently, in this case a multi-modal traffic simulation is conducted and micromobility will affect private cars traffic stream. Case
3) If a possible cycling route does not exist in the PSRC road network, we assume short vehicle trips are not able to be replaced by
micromobility. As a result, no changes are made to the origin–destination matrix.
Elevation is also a crucial factor that affects cycling, especially in hilly urban areas. One of the perceived reasons for the failure of
Pronto – Seattle’s first bike-sharing system is the city’s challenging terrain (Sun et al., 2018). In 2020, Seattle deployed a shared pedal
assist electric bike (e-bike) system to overcome elevation barriers and substantially reduce the energy and effort needed to bike on
uphill segments. For example, pedal assist e-bikes require 17% less effort than a regular bike on uphill segments (Langford et al., 2017),
enabling users to take longer trips, even on hilly routes. As a result, we do not consider any elevation limitations in the traffic
simulation model and assume that micromobility electrification can help to overcome the terrain challenges in the city.

5.3. Comparison metrics

To allow an evaluation of the system and link-level impacts of replacing short car trips with micromobility modes, scenarios are
evaluated with respect to four key indicators:

• Congestion: travel time and speed changes at the link-level and all travel paths within the whole network. Specifically, we measure
speed by the ratio of travel speed over the free flow speed of the link, where free-flow speed is defined as the mean speed of
passenger cars under low to moderate flow rates that can be accommodated on a uniform roadway under prevailing roadway and
traffic conditions (TRB Highway Capacity Manual, 2016)
• Total Travel Demand: measured in terms of VMT and is the sum of the travel distances of all trips in the traffic simulation model.
• Emissions: the distance of all trips multiplied by an emissions factor. The method for estimating total system-wide GHG emissions is
defined below in Equation (5) (adapted from Harper et al., 2018), expressed as
∑
GHG = (CO2 /gal)∙(gal/mile)∙VMTI (5)
i

where GHG = total GHG emissions emitted from light-duty vehicles, CO2 /gal = amount of direct CO2 for one gallon of gasoline (8,890
g) (National Highway Traffic Safety Administration, 2010), gal/mile = average fuel consumption for passenger cars in the city of
Seattle for the year 2014 (1 Gal/23 miles) (Seattle Office of Sustainability and Environment, 2016), andVMT i = the amount of VMT
generated by light-duty vehicle i.

• Energy Use: We will also estimate system-level energy consumption following a similar process to estimate emissions. The equation
used to estimate energy use is defined in Equation (6), expressed as
∑
E= (MJ/gal)∙(gal/mile)∙I (6)
i

where, E = total energy consumed by light-duty vehicles, MJ/gal = the amount of energy in a gallon of gasoline (127.1 MJ/gal) (EIA
2016) (U.S. Energy Information Administration, 2021).
Metrics are compared between the base case scenario and updated scenarios where a portion of private vehicle trips are replaced
with micromobility modes. Literature shows that micromobility modes produce GHG emissions and consume energy over their life
cycle (Reck et al., 2021) (Luo et al., 2020) (de Bortoli and Christoforou, 2020) (Hollingsworth et al., 2019). Therefore, in updated
scenarios, we use estimates of life-cycle emissions for shared e-bikes per VMT for the fuel and operational services life cycle com­
ponents from ITF (International Transport Forum, 2020) to estimate net savings from replacing short car trips with micromobility
modes. Specifically, we consider fuel component emissions and energy use due to e-bike charging, and operational services emissions
and energy use from shared bike system rebalancing. Net energy use and emissions were found using a similar method to (Reck et al.,
2021), where we take the difference between the emissions that would have been generated if the replaced trips were done by a private
vehicle and the emissions that would have been generated if these same trips were done by shared e-bike. A positive net emissions
estimate can be interpreted as additional emissions caused by the new mode, whereas negative emissions can be interpreted as
emissions saved by the new mode.

10
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

6. Results

In order to assess the environmental and congestion benefits of higher micromobility penetration, we simulate 3–4 pm peak hour
under normal conditions (base case) and assess improvements in link performance as we replace the short vehicle trips by
micromobility.

6.1. Seattle traffic congestion alleviation

To evaluate the congestion implications of short car trip substitution with micromobility, we first examine the performance of links
relative to free flow speed. It is important to evaluate this ratio as this is a direct indicator of link performance and traffic condition,
which is not captured directly from observing absolute changes in travel speed. For example, even if a link speed increases by 10% or 5
mph, without considering free flow speed or speed limit, it is difficult to assess whether link performance is now satisfactory.
Fig. 6 depicts the percent change in congested links at different micromobility penetration rates when compared to the base case
scenario, where gray scale bars demonstrate the 2014 bike-infrastructure situation where in many cases bikes must share roads with
vehicles and affect the traffic stream; the orange bars is a future scenario where protected bike lanes are deployed for every short trip
origin–destination pair. In the base case scenario, there are 73 links with severe congestion, 490 links with moderate congestion, 2,285
links with mild congestion, and 6,792 links with no congestion. As expected, when the micromobility penetration rate increases, a
larger portion of links become less congested. At the 10% penetration rate, the number of links with moderate and severe congestion
decreases by about 3% and 1.5%, respectively, while the number of non-congested links increase by about 0.5%. At the upper bound
penetration rate (18%), non-congested links increase about 1%, the number of moderately and severely congested links decrease by
around 3% and 10%, respectively. If protected bike lanes were available between all short trip origin–destination pairs, the congestion
alleviation effect could be doubled. For example, at the upper bound penetration rate, the percent change in the number of non-
congested links is twice as many as that of the 2014 bike lane configuration scenario. It should be noted that the improvements in
traffic performance does not mean all congested roads have converted to non-congested, rather, the overall urban traffic situation has
been improved and those links that are under lower congestion levels may become free-flow roads. Although the simulation shows
there are some congestion benefits from replacing short car trips with micromobility, these trips still affect the private vehicle stream
due to the sparse deployment of dedicated bicycle infrastructure throughout the city, indicating that wide-scale deployment of bike
lanes is needed to maximize congestion benefits.

6.2. Link-level travel speed and time improvement

Beyond assessing the number of links under various congestion levels with micromobility substitution, we also examine the ab­
solute changes in travel speed for every link at different penetration rates when compared to the base case scenario (Fig. 7). As we
increase the micromobility penetration rate, we observe improvements in terms of travel speed, as represented by the darker shades of
blue. At 10% penetration rate, we see that speeds on most roadways increase between 0% and 1%. At the upper bound penetration rate

Fig. 6. Percent Change in Number of Congested Links under Different Micromobility Penetration Rates. a Wide-scale bike lane deployment scenario.
It assumes that will be a protected bike lane available between every origin–destination TAZ pair. b 2014 bike lane configuration scenario. which
uses the bike lane configuration from the Puget Sound Regional Council network model. Note: Links where the ratio of current speed to free flow
speed is greater than or equal to 0.95, is defined as “No Congestion”; 0.8–0.95 as “Mild Congestion”; 0.6–0.8 as “Moderate Congestion”; and less
than 0.6 as “Severe Congestion”.

11
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

improvements are more noticeable (1% − 5%) at roads that are more congested at the base scenario, particularly at urban arterial
intersections where traffic is more likely to be clogged by traffic signals or stop/yield signs or on freeways/highways in areas where
there are bottlenecks. More significant impact on street intersections (more than 5%) can be observed if bike lanes are sufficiently
deployed.
The results shows that as more short car trips are replaced by micromobility, the severely congested roads in the base scenario have
more opportunity to be relieved. The reason why such phenomenon appears is that the more heavily congested a roadway is, the more
sensitive to changes in volume. The change in speed for each link is based on the volume delay function, which provides a larger
reward for removing cars from the more heavily congested roadways when compares to less congested roads. Although overall traffic
conditions improve, there are some roads that experience a decrease in speed. This is because agents are choosing to take other routes
that have less cars to satisfy the user’s equilibrium assignment. When some short car trips are replaced with micromobility, less de­
mand is assigned to the traffic network, a new equilibrium is obtained and those roadways experiencing severe congestion such as
urban arterial street intersections and bridges experience the greatest improvements.
Fig. 8 displays the average travel speed and time improvement at link level by facility type (i.e., urban arterials and expressway/
freeway) and congestion level for different micromobility penetration rates when compared to the base case scenario. The four sub-
graphs show that urban arterials tend to benefit more from short car trip replacement than do freeways/expressways, in terms of both
link travel time and speed. For example, travel speeds on severely congested freeway and urban arterial links increase by about 0.6%
and 1%, respectively, at the upper bound penetration rate. This is due to the fact that most short vehicle trips occur on urban arterials
and as a result these facilities experience greater benefits as short car trips are removed from the road network and replaced with
micromobility modes. Greater speed and travel time benefits can be achieved as more dedicated bike lanes are deployed.

6.3. Impacts to vehicle miles traveled, emissions, and energy use

On-road GHGs comprise about 40% of Seattle’s total emissions. In 2014, the annual emissions from the road transportation sector
totaled about 2.34 million metric tons of CO2, with passenger cars comprising about 75% or 1.72 million metric tons CO2 of all surface
transportation emissions in the city (Seattle Office of Sustainability and Environment, 2016). In order to estimate average peak hour
weekday VMT for the city of Seattle, we use the static traffic assignment model described in the Method section. The base VMT in this
study was estimated using a similar process to that found in the 2014 Seattle Community Greenhouse Gas Emissions Inventory (Seattle
Office of Sustainability and Environment, 2016). This study as well as the inventory employs a method that counts emissions from all
trips that occur entirely within Seattle, half of the trips that either begin or end in the city, and no trips that both begin and end outside
the city (even if they pass through the city), known as an origin–destination pair approach. This is an increasingly common way of

Fig. 7. Percent Change in Travel Speed under Different Micromobility Penetration Rates.

12
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

Fig. 8. Average Percent Change in Travel Time and Speed by Facility Type and Penetration Rate for Road Network in Seattle.

counting GHG emissions in community-scale inventories and was recommended in the International Council for Local Environmental
Initiatives’ US Community Protocol (Local Governments for Sustainability, 2019). We estimate that total light-duty VMT in Seattle
during the 3–4 PM peak hour is about 0.68 million miles.
Micromobility modes are not carbon neutral and produce life-cycle emissions. Different micromobility types differ in terms of
emissions and energy characteristics, as discussed in the Methods section. We follow use stage life cycle estimates from ITF

Table 1
Emissions and Energy Savings in 3–4 PM Peak Hours.
Micromobility Number of Short Percent of Total Total Daily Total Net Savings in Total Net Savings in Percent Net Change in
Penetration Rate Trips Replaced Trips Replaced Decrease in Daily Emissions (metric Daily Energy Use Emissions and Energy
VMT tons CO2) (GJ) Use

10% 4,866 4.56% 7,359 − 2.76 − 39.34 − 1.05%

18% 8,833 8.26% 13,696 − 5.14 − 73.29 − 1.97%

Note: VMT = Vehicle Miles Traveled.

Note: This table focuses on those trips that originated and/or ended in Seattle.
Note: Shared e-bike is the representative micromobility mode for the environmental impacts estimation. Assuming all trips at the upper bound
penetration rate are done by other types of micromobility vehicles, percent net savings in emissions and energy use are as follows: shared e-scooter:
− 1.97%, shared bike: − 1.99%, private e-bike: − 2.01%, private e-scooter: − 2.02%, private bike: − 2.03%.

13
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

(International Transport Forum, 2020) to estimate the net savings in emissions and energy use if a portion of short car trips were
replaced with micromobilty modes. Given that Seattle has deployed a fleet of e-bikes, which are capable of handling the city’s hilly
terrain, shared e-bikes are selected as the representative micromobility mode for our environmental impacts assessment. Emissions
from charging and rebalancing shared e-bikes are compared to the emissions from driving a car for the same trips.
Table 1 shows the changes in emissions and energy use if short car trips were replaced with micromobility modes under different
penetration rates. In this study, we estimate changes in light-duty VMT and emissions, and energy use can be obtained according to
VMT. It is shown that at the upper bound penetration rate (18%), switching to micromobility modes from short car trips could reduce
light-duty transportation emissions and energy use by about 2%. This is equivalent to 5 tons carbon dioxide (CO2), and 73 Giga Joules
(GJ) per workday during 3-4PM peak hours, which translates to 1,130 tons of CO2 and 16,124 GJ per year, assuming there are 220
workdays in a calendar year. At the upper bound penetration rate shared e-bikes produce 0.15 metric tons CO2, whereas light-duty
vehicles produce 5.29 metric tons CO2 for the same trips.

7. Conclusion

In this paper, we explored the potential impacts to congestion, emissions, and energy use if a proportion of short car trips are
replaced by micromobility modes, using Seattle as a case study. To conduct this analysis, we use a combination of travel demand, travel
survey, and weather data. Currently about 48% of peak hour trips in Seattle are short trips (0–3 miles). Approximately 18% of short
trips in Seattle can be replaced by micromobility modes, when considering commuter age, tour and trip purposes, time of day, and
weather limitations that prevent micromobility from being a viable alternative for certain private car trips. The upper bound potential
is not uniformly distributed across the case study area and it is shown that downtown area has a higher potential that other parts of the
city. A recommendation for cities to increase the upper bound penetration rate is to deploy cargo bikes with the necessary storage
capacity for residents to be able to substitute grocery shopping trips done by personal vehicle with e-bikes. Such a measure being
implemented in Seattle would increase the upper bound penetration rate to 29%, following the same framework.
In order to simulate peak hour traffic under current conditions and scenarios where short trips are replaced by micromobility
modes, we develop a static traffic assignment model based on PSRC’s demand and network model and calibrate the model according to
observed road traffic count volumes. To assess benefits, we compare current (i.e., 2014 base case) congestion conditions and travel
speed results to micromobility penetration rate scenarios. The results show that under the upper bound micromobility penetration rate,
there are around 10% less severely congested road segments and 1% more non-congested road segments, when compared to our base
case scenario. Traffic simulation also reveals that urban arterials benefit much more than freeways/expressways with micromobility
substitution. This is because short car trips are more likely to occur on urban arterials rather than freeways and highways. Moreover, no
matter the facility type, the more congested roads tend to experience greater improvements in performance, meaning the strategy of
replacing short private vehicle trips with micromobility is most effective on congested road segments. We also observed that if pro­
tected bike lanes were available between each short trip origin–destination pair, benefits to congestion in terms of the percent change
in the number of non-congested links can be doubled at the upper bound penetration rate, when compared to the 2014 bike lane
configuration. We also see more significant increases in travel speed under the wide-scale bike-lane deployment scenario on most
roadways. Results indicate that replacing short private vehicle trips with micromobility is a feasible solution to mitigate traffic
congestion, especially on heavily congested links such as bridges and urban arterial street intersections.
At the upper bound micromobility penetration rate (i.e., 18%), we estimate about a 2% net reduction in light-duty emissions and
energy use, for the 3–4 pm peak hour simulated. This suggests that while micromobility could have significant impacts on congestion,
the impacts to energy use and emissions are disproportionately low when compared to the percent of trips replaced. This is because
short trips do not generate a large amount of VMT per trip, so the percent change on trips is not equivalent to the percent change in
emissions or energy use. This would also indicate that while micromobility is green alternative to driving, other measures are needed to
help cities substantially reduce their transportation sector emissions and meet climate targets set forth by the White House (The White
House, 2021). For example, cities can target long trips, which make up the bulk of peak hour emissions, by expanding commuter transit
options (e.g., commuter rail and bus). This will provide residents living outside of the city with a viable and reliable alternative to get to
and from work during peak hour travel. Transit agencies should phase out the existing bus fleet with zero-emissions buses and truck
companies should deploy electric trucks as the current heavy-duty vehicle fleet phases out. Incentives for light-duty electric vehicles
(EVs) should remain in place at the federal, state, and local levels to encourage car owners with internal combustion engine vehicles to
switch to EVs. Finally, last-mile parcel deliveries done traditionally by step vans could be replaced with micromobility modes, when
possible.
In order to evaluate the feasibility for micromobility to replace short private vehicle trips, we assess the extent to which the current
shared micromobility system in Seattle could satisfy the estimated upper bound number of trips. In May 2021 there were about 500
total trips done by shared bikes and scooters collectively during the 3–4 pm peak hour, whereas there were a total of 5,300 shared
micromobility vehicles (1,700 bikes and 3,600 scooters) deployed in Seattle during this timeframe (Seattle Department of Trans­
portation, 2021). While there may need to be more bikes shared micromobility vehicles deployed to satisfy the upper bound demand
estimated in this paper (i.e., 8,800 trips), this also indicates that the vast majority of shared micromobility vehicles are not in use at any
one time during the day. To encourage micromobility mode usage over private cars we make the following recommendations: 1) public
agencies in conjunction with micromobility providers should seek to establish economic incentives such as providing free bike rides for
transit riders, similar to the partnership between Healthy ride bikeshare company and the Port Authority of Allegheny County (Healthy
Ride, 2021); 2) municipalities may consider implementing congestion fees for drivers who wish to enter the downtown area; and 3)
bikeshare providers should distribute bikeshare stations equitably across the region so that a greater number of residents, including

14
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

low-income and minority populations also have access to shared micromobility modes.
While the results of this study offer a new understanding of the congestion and environmental impacts of short car trip replacement
with micromobility modes, there are several opportunities for model improvement and future work. The upper bound micromobility
penetration rate estimation is based on a survey conducted during the Spring 2014 season. It is known that cycling is subject to weather
and location (Noland, 2021) (Miranda-Moreno and Nosal, 2011), so if other seasons are taken into consideration, the upper bound
estimation may be different - during winter there will be days where freezing temperatures and snow conditions pose as a restriction to
micromobility travel. For instance, research shows that both dockless or docked bike-sharing systems receive much less ridership in
rainier and colder months than warmer and drier months in Seattle (Peters and MacKenzie, 2019). Therefore, the environmental
impacts calculation across the year needs further investigation. Moreover, if apply our methodology to other cities around the world,
the upper bound may be different because of climates. This study assesses the congestion effects of replacing short car trips with
micromobility modes by using passenger car equivalent guidelines from the Highway Capacity Manual. Future studies could conduct a
more detailed analysis using methods such as field testing to assess how micromobility modes sharing roads with cars affects travel
speed. Finally, the emissions and energy consumption estimates shown in this paper are based on changes in VMT and do not consider
the impacts of improvements in vehicle speed across the system on these estimates. Future studies should take this into account to
develop more robust estimates of changes in emissions and energy use that come from replacing short car trips with micromobility
modes.

CRediT authorship contribution statement

Zhufeng Fan: Formal analysis, Data curation, Software, Visualization, Writing – original draft. Corey D. Harper: Conceptuali­
zation, Methodology, Funding acquisition, Investigation, Supervision, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Acknowledgements

This work was supported by a USDOT University Transportation Center [Grant Number DTRT12GUTC11], by a grant from the
Commonwealth of Pennsylvania, Department of Community and Econmic Development, and the Hillman Foundation’s support of the
Traffic 21 Institute at Carnegie Mellon University. We sincerely appreciate Kris Overby, Stefan Coe, and Brice Nichols from PSRC for
providing the travel demand model data, and their continuous help for our baseline model establishment.

References

American Meteorological Society, 2015. Beaufort wind scale - Glossary of Meteorology [WWW Document]. URL https://glossary.ametsoc.org/wiki/Beaufort_wind_
scale (accessed 10.30.21).
Belloli, M., Giappino, S., Robustelli, F., Somaschini, C., 2016. Drafting Effect in Cycling: Investigation by Wind Tunnel Tests. Procedia Eng. Eng. SPORT 11 (147),
38–43. https://doi.org/10.1016/j.proeng.2016.06.186.
Brand, C., Dons, E., Anaya-Boig, E., Avila-Palencia, I., Clark, A., de Nazelle, A., Gascon, M., Gaupp-Berghausen, M., Gerike, R., Götschi, T., Iacorossi, F., Kahlmeier, S.,
Laeremans, M., Nieuwenhuijsen, M.J., Pablo Orjuela, J., Racioppi, F., Raser, E., Rojas-Rueda, D., Standaert, A., Stigell, E., Sulikova, S., Wegener, S., Int Panis, L.,
2021. The climate change mitigation effects of daily active travel in cities. Transport. Res. Part D Transport Environ. 93 https://doi.org/10.1016/j.
trd.2021.102764.
Carmichael, C., 2020. Cycling in Wind: How to Ride Smart, Fast, and Strong. URL https://trainright.com/cycling-in-wind/ (accessed 6.1.21).
Cervero, R., 2003. Road Expansion, Urban Growth, and Induced Travel: A Path Analysis. J. American Planning Assoc. 69, 145–163. https://doi.org/10.1080/
01944360308976303.
Cherry, C.R., Weinert, J.X., Xinmiao, Y., 2009. Comparative environmental impacts of electric bikes in China. Transport. Res. Part D Transport Environ. 14, 281–290.
https://doi.org/10.1016/j.trd.2008.11.003.
Church, P.E., 1974. Some Precipitation Characteristics of Seattle. Weatherwise 27, 244–251. https://doi.org/10.1080/00431672.1974.9931720.
Darksky, 2021. Darksky [WWW Document]. URL https://darksky.net/forecast/40.7127,-74.0059/us12/en (accessed 4.1.21).
de Bortoli, A., Christoforou, Z., 2020. Consequential LCA for territorial and multimodal transportation policies: method and application to the free-floating e-scooter
disruption in Paris. J. Cleaner Prod. 273 https://doi.org/10.1016/j.jclepro.2020.122898.
Fan, Y., Zheng, S., 2020. Dockless bike sharing alleviates road congestion by complementing subway travel: Evidence from Beijing. Cities 107. https://doi.org/
10.1016/j.cities.2020.102895.
Fishman, E., Washington, S., Haworth, N., 2014. Bike share’s impact on car use: Evidence from the United States, Great Britain, and Australia. Transport. Res. Part D
Transport Environ. 31, 13–20. https://doi.org/10.1016/j.trd.2014.05.013.
Gebhart, K., Noland, R.B., 2014. The impact of weather conditions on bikeshare trips in Washington, DC. Transportation 41, 1205–1225. https://doi.org/10.1007/
s11116-014-9540-7.
Hamilton, T.L., Wichman, C.J., 2018. Bicycle infrastructure and traffic congestion: Evidence from DC’s Capital Bikeshare. J. Environ. Economics Manage. 87, 72–93.
https://doi.org/10.1016/j.jeem.2017.03.007.
Harper, C.D., Hendrickson, C.T., Samaras, C., 2018. Exploring the Economic, Environmental, and Travel Implications of Changes in Parking Choices due to Driverless
Vehicles: An Agent-Based Simulation Approach. J. Urban Plann. Dev. 144, 04018043. https://doi.org/10.1061/(ASCE)UP.1943-5444.0000488.
Healthy Ride, 2021. Healthy Ride Pittsburgh | Port Authority ConnectCardHealthy Ride Pittsburgh [WWW Document]. URL https://healthyridepgh.com/
connectcard/ (accessed 11.19.21).
Hollingsworth, J., Copeland, B., Johnson, J.X., 2019. Are e-scooters polluters? The environmental impacts of shared dockless electric scooters. Environ. Res. Lett. 14
https://doi.org/10.1088/1748-9326/ab2da8.
INRIX, 2020. 2020 Global Traffic Scorecard [WWW Document]. URL https://inrix.com/scorecard/ (accessed 3.31.21).

15
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

INRIX, 2019. Shared Bikes and Scooters Could Replace Nearly 50 Percent of Downtown Vehicle Trips [WWW Document]. URL https://inrix.com/press-releases/
micromobility-study-us-2019/ (accessed 3.30.21).
International Transport Forum, 2020. Good to Go? Assessing the Environmental Performance of New Mobility 89.
Isvan, O., 2015. Wind speed, wind yaw and the aerodynamic drag acting on a bicycle and rider. J. Sci. Cycl 4, 42–50.
Ji, S., Cherry, C.R., J. Bechle, M., Wu, Y., Marshall, J.D., 2012. Electric Vehicles in China: Emissions and Health Impacts. Environ. Sci. Technol. 46, 2018–2024.
https://doi.org/10.1021/es202347q.
Johnson, M., Rose, G., 2015. Extending life on the bike: Electric bike use by older Australians. J. Transport Health 2, 276–283. https://doi.org/10.1016/j.
jth.2015.03.001.
Kou, Z., Wang, X., Chiu, S.F. (Anthony), Cai, H., 2020. Quantifying greenhouse gas emissions reduction from bike share systems: a model considering real-world trips
and transportation mode choice patterns. Resources, Conservation and Recycling 153, 104534. https://doi.org/10.1016/j.resconrec.2019.104534.
Langford, B.C., Cherry, C.R., Bassett, D.R., Fitzhugh, E.C., Dhakal, N., 2017. Comparing physical activity of pedal-assist electric bikes with walking and conventional
bicycles. J. Transport Health 6, 463–473. https://doi.org/10.1016/j.jth.2017.06.002.
Li, A., Gao, K., Zhao, P., Qu, X., Axhausen, K.W., 2021a. High-resolution assessment of environmental benefits of dockless bike-sharing systems based on transaction
data. J. Cleaner Prod. 296 https://doi.org/10.1016/j.jclepro.2021.126423.
Lime, 2021. How old do I have to be to ride Lime? [WWW Document]. URL https://help.li.me/hc/en-us/articles/115004915588-How-old-do-I-have-to-be-to-ride-
Lime- (accessed 6.1.21).
Local Governments for Sustainability, 2019. U.S. Community Protocol for Accounting and Reporting of Greenhouse Gas Emissions.
Lovelace, R., Beck, S.B.M., Watson, M., Wild, A., 2011. Assessing the energy implications of replacing car trips with bicycle trips in Sheffield, UK. Energy Policy 39,
2075–2087. https://doi.org/10.1016/j.enpol.2011.01.051.
Luo, H., Zhao, F., Chen, W.-Q., Cai, H., 2020. Optimizing bike sharing systems from the life cycle greenhouse gas emissions perspective. Transport. Res. Part C Emerg.
Technolog. 117 https://doi.org/10.1016/j.trc.2020.102705.
Martin, R., 2021. The future of micromobility: How short, shared rides are disrupting car travel [WWW Document]. URL https://www.thezebra.com/resources/
driving/micromobility/ (accessed 5.15.21).
Mason, J., Fulton, L., McDonald, Z., 2015. A Global High Shift Cycling Scenario: The Potential for Dramatically Increasing Bicycle and E-bike Use in Cities Around the
World, with Estimated Energy, CO2, and Cost Impacts.
Matheny, F., 2019. Win Against Wind: A Guide to Cycling in Windy Conditions [WWW Document]. Road Bike Rider Cycling Site. URL https://www.roadbikerider.
com/ultimate-guide-cycling-wind/ (accessed 6.1.21).
McQueen, M., MacArthur, J., Cherry, C., 2020. The E-Bike Potential: Estimating regional e-bike impacts on greenhouse gas emissions. Transport. Res. Part D Transport
Environ. 87 https://doi.org/10.1016/j.trd.2020.102482.
Merten, P., 2020. Study: Seattle-area congestion carries $1.8B economic impact. Puget Sound Business J.
Miranda-Moreno, L.F., Nosal, T., 2011. Weather or Not to Cycle: Temporal Trends and Impact of Weather on Cycling in an Urban Environment. Transp. Res. Rec.
2247, 42–52. https://doi.org/10.3141/2247-06.
Mobike, 2018a. How Cycling Changes Cities – Mobike’s Second Whitepaper Goes Global [WWW Document]. URL https://mobike.com/sg/blog/post/cycling-changes-
cities (accessed 8.4.21).
Mobike, 2018b. Terms and Conditions [WWW Document]. URL https://mobike.com/sg/terms (accessed 6.1.21).
Nankervis, M., 1999. The effect of weather and climate on bicycle commuting. Transport. Res. Part A Policy Pract. 33, 417–431. https://doi.org/10.1016/S0965-8564
(98)00022-6.
National Association of City Transportation Officials, 2020. 136 Million Trips Taken on Shared Bikes and Scooters Across the U.S. in 2019.
National Highway Traffic Safety Administration, 2010. National Highway Traffic Safety Administration [WWW Document]. URL https://www.nhtsa.gov/ (accessed
5.6.21).
Noland, R., Ishaque, M., 2006. Smart Bicycles in an Urban Area: Evaluation of a Pilot Scheme in London. J. Public Transportat. 9 https://doi.org/10.5038/2375-
0901.9.5.5.
Noland, R.B., 2021. Scootin’ in the rain: Does weather affect micromobility? Transportat. Res. Part A Policy Pract. 149, 114–123. https://doi.org/10.1016/j.
tra.2021.05.003.
Peters, L., MacKenzie, D., 2019. The death and rebirth of bikesharing in Seattle: Implications for policy and system design. Transportat. Res. Part A Policy Pract. 130,
208–226. https://doi.org/10.1016/j.tra.2019.09.012.
Pucher, J., Buehler, R., Seinen, M., 2011. Bicycling renaissance in North America? An update and re-appraisal of cycling trends and policies. Transport. Res. Part A
Policy Pract. 45, 451–475. https://doi.org/10.1016/j.tra.2011.03.001.
Puget Sound Regional Council, 2016. Spring 2014 Household Survey [WWW Document]. URL https://www.psrc.org/travel-surveys-spring-2014-household-survey
(accessed 6.2.21).
Reck, D.J., Martin, H., Axhausen, K.W., 2021. Mode choice, substitution patterns and environmental impacts of shared and personal micro-mobility. Presented at the
21st Swiss Transport Research Conference (STRC 2021), STRC. https://doi.org/10.3929/ethz-b-000504162.
Reed, T., 2019. Micromobility Potential in the US, UK and Germany. INRIX.
Rojas-Rueda, D., de Nazelle, A., Teixidó, O., Nieuwenhuijsen, M.J., 2012. Replacing car trips by increasing bike and public transport in the greater Barcelona
metropolitan area: A health impact assessment study. Environ. Int. 49, 100–109. https://doi.org/10.1016/j.envint.2012.08.009.
Ryan Blethen, 2019. Seattle-area traffic congestion is among the worst in the country, study shows. The Seattle Times.
Saneinejad, S., Roorda, M.J., Kennedy, C., 2012. Modelling the impact of weather conditions on active transportation travel behaviour. Transportat. Res. Part D
Transport Environ. 17, 129–137. https://doi.org/10.1016/j.trd.2011.09.005.
Schmiedeskamp, P., Zhao, W., 2016. Estimating Daily Bicycle Counts in Seattle, Washington, from Seasonal and Weather Factors. Transp. Res. Rec. 2593, 94–102.
https://doi.org/10.3141/2593-12.
Schrank, D., Eisele, B., Lomax, T., 2019. 2019 Urban Mobility Report.
Sears, J., Flynn, B.S., Aultman-Hall, L., Dana, G.S., 2012. To Bike or Not to Bike: Seasonal Factors for Bicycle Commuting. Transp. Res. Rec. 2314, 105–111. https://
doi.org/10.3141/2314-14.
Seattle Department of Transportation, 2021. Bike Share - Transportation | seattle.gov [WWW Document]. URL http://www.seattle.gov/transportation/projects-and-
programs/programs/bike-program/bike-share (accessed 11.19.21).
Seattle DOT, 2021. New Mobility Program [WWW Document]. URL https://www.seattle.gov/transportation/projects-and-programs/programs/new-mobility-
program (accessed 5.17.21).
Seattle GeoData Portal, 2019. 2014 Traffic Flow Counts [WWW Document]. URL https://data-seattlecitygis.opendata.arcgis.com/datasets/2014-traffic-flow-counts
(accessed 7.19.21).
Seattle Office of Sustainability and Environment, 2016. Seattle Office of Sustainability and Environment [WWW Document]. URL https://www.seattle.gov/
environment (accessed 7.28.21).
Shaheen, S.A., Guzman, S., Zhang, H., 2010. Bikesharing in Europe, the Americas, and Asia: Past, Present, and Future. Transp. Res. Rec. 2143, 159–167. https://doi.
org/10.3141/2143-20.
Shalini, K., 2014. Estimation of the Passenger Car Equivalent: A Review. Int. J. Emerg. Technol. Adv. Eng. 4, 97–102.
StoIlova, K., StoIlov, T., 1998. Traffic noise and traffic light control. Transportat. Res. Part D Transport Environ. 3, 399–417. https://doi.org/10.1016/S1361-9209
(98)00017-0.
Sun, F., Chen, P., Jiao, J., 2018. Promoting public bike-sharing: A lesson from the unsuccessful Pronto system. Transport. Res. Part D Transport Environ. 63, 533–547.
https://doi.org/10.1016/j.trd.2018.06.021.

16
Z. Fan and C.D. Harper Transportation Research Part D 103 (2022) 103173

The White House, 2021. FACT SHEET: President Biden’s Leaders Summit on Climate [WWW Document]. The White House. URL https://www.whitehouse.gov/
briefing-room/statements-releases/2021/04/23/fact-sheet-president-bidens-leaders-summit-on-climate/ (accessed 8.3.21).
The World Bank, 2021. Urban population - United States [WWW Document]. URL https://data.worldbank.org/indicator/SP.URB.TOTL?
end=2019&locations=US&start=2001 (accessed 3.30.21).
TRB Highway Capacity Manual, 2016. Highway Capacity Manual, Sixth Edition.
Trenberth, K.E., Zhang, Y., 2018. How Often Does It Really Rain? Bull. Am. Meteorol. Soc. 99, 289–298. https://doi.org/10.1175/BAMS-D-17-0107.1.
United Nations Department of Economic and Social Affairs, 2014. 2014 revision of the World Urbanization Prospects [WWW Document]. URL 2014-revision-world-
urbanization-prospects.html (accessed 3.30.21).
US Bureau of Transportation Statistics, 2020. Bikeshare and e-scooters in the U.S. [WWW Document]. URL https://data.bts.gov/stories/s/Bikeshare-and-e-scooters-in-
the-U-S-/fwcs-jprj/ (accessed 3.31.21).
US Census Bureau, 2021. Public Use Microdata Areas (PUMAs) [WWW Document]. Census.gov. URL https://www.census.gov/programs-surveys/geography/
guidance/geo-areas/pumas.html (accessed 10.31.21).
U.S. Energy Information Administration [WWW Document], 2021. URL https://www.eia.gov/index.php (accessed 5.6.21).
USDOT Federal Highway Administration, 2021. Office of Highway Policy Information [WWW Document]. Office of Highway Policy Information. URL https://www.
fhwa.dot.gov/policyinformation/travel_monitoring/tvt.cfm (accessed 3.30.21).
Van Cauwenberg, J., De Bourdeaudhuij, I., Clarys, P., de Geus, B., Deforche, B., 2019. E-bikes among older adults: benefits, disadvantages, usage and crash
characteristics. Transportation 46, 2151–2172. https://doi.org/10.1007/s11116-018-9919-y.
Wang, M., Zhou, X., 2017. Bike-sharing systems and congestion: Evidence from US cities. J. Transp. Geogr. 65, 147–154. https://doi.org/10.1016/j.
jtrangeo.2017.10.022.
Winslott Hiselius, L., Svensson, Å., 2017. E-bike use in Sweden – CO2 effects due to modal change and municipal promotion strategies. J. Cleaner Prod. 141, 818–824.
https://doi.org/10.1016/j.jclepro.2016.09.141.
WSDOT, 2021. Washington State bicycle laws [WWW Document]. URL https://wsdot.wa.gov/travel/commute-choices/bike/laws (accessed 10.30.21).
Zahabi, S.A.H., Chang, A., Miranda-Moreno, L.F., Patterson, Z., 2016. Exploring the link between the neighborhood typologies, bicycle infrastructure and commuting
cycling over time and the potential impact on commuter GHG emissions. Transportat. Res. Part D Transport Environ. 47, 89–103. https://doi.org/10.1016/j.
trd.2016.05.008.
Zhang, Y., Mi, Z., 2018. Environmental benefits of bike sharing: A big data-based analysis. Appl. Energy 220, 296–301. https://doi.org/10.1016/j.
apenergy.2018.03.101.
Zhao, J., Guo, C., Zhang, R., Guo, D., Palmer, M., 2019. Impacts of weather on cycling and walking on twin trails in Seattle. Transportat. Res. Part D Transport
Environ. 77, 573–588. https://doi.org/10.1016/j.trd.2019.09.022.

17