LETTER • OPEN ACCESS

Are e-scooters polluters? The environmentalimpacts of shared dockless electric scootersTo cite this article: Joseph Hollingsworth et al 2019 Environ. Res. Lett. 14 084031

 

View the article online for updates and enhancements.

You may also likeSoftware development of electric scooterD G Sesa Balaji, S Shabeer Ahamed andD Kirubakaran

-

Analysis of an employment of a gear ratiorate in CVT control systemW Grzegoek and M Szczepka

-

Evaluation of mitigation effects on airpollutants for electric scooters in Taiwanwith the energy flow analysis and systemdynamics approachP Y Hsieh, L F W Chang, T Y Yu et al.

-

Recent citationsLife Cycle Assessment on Electric MopedScooter SharingNora Schelte et al

-

Institutional Logics at Play in a Mobility-as-a-Service EcosystemHugo Guyader et al

-

How Do Shared Dockless E-ScooterServices Affect Mobility Practices in Paris?A Survey-Based Estimation of Modal ShiftCamille Krier et al

-

This content was downloaded from IP address 65.21.228.167 on 20/10/2021 at 10:44

Environ. Res. Lett. 14 (2019) 084031 https://doi.org/10.1088/1748-9326/ab2da8

LETTER

Are e-scooters polluters? The environmental impacts of shareddockless electric scooters

JosephHollingsworth, BrennaCopeland and JeremiahX Johnson1

Department of Civil, Construction, and Environmental Engineering, NorthCarolina StateUniversity, United States of America1 Author towhomany correspondence should be addressed.

E-mail: jjohns24@ncsu.edu

Keywords: electric scooter, life cycle assessment, transportation, environmental impacts

Supplementarymaterial for this article is available online

AbstractShared stand-up electric scooters are nowoffered inmany cities as an option for short-term rental, andmarketed for short-distance travel. Using life cycle assessment, we quantify the total environmentalimpacts of thismobility option associatedwith global warming, acidification, eutrophication, andrespiratory impacts.Wefind that environmental burdens associatedwith charging the e-scooter aresmall relative tomaterials andmanufacturing burdens of the e-scooters and the impacts associatedwith transporting the scooters to overnight charging stations. The results of aMonteCarlo analysisshow an average value of life cycle global warming impacts of 202 gCO2-eq/passenger-mile, driven bymaterials andmanufacturing (50%), followed by daily collection for charging (43%of impact).Weillustrate the potential to reduce life cycle global warming impacts through improved scootercollection and charging approaches, including the use of fuel-efficient vehicles for collection (yielding177 gCO2-eq/passenger-mile), limiting scooter collection to thosewith a lowbattery state of charge(164 gCO2-eq/passenger-mile), and reducing the driving distance per scooter for e-scooter collectionand distribution (147 gCO2-eq/passenger-mile). The results prove to be highly sensitive to e-scooterlifetime; ensuring that the shared e-scooters are used for two years decreases the average life cycleemissions to 141 gCO2-eq/passenger-mile. Under our BaseCase assumptions, wefind that the lifecycle greenhouse gas emissions associatedwith e-scooter use is higher in 65%of ourMonteCarlosimulations than the suite ofmodes of transportation that are displaced. This likelihood drops to35%–50%under our improved and efficient e-scooter collection processes and only 4%whenweassume two-year e-scooter lifetimes.When e-scooter usage replaces average personal automobiletravel, we nearly universally realize a net reduction in environmental impacts.

1. Introduction

With a small electric motor and a deck on which asingle rider stands, stand-up scooters are designed totransport riders short distances around urban settings.Ride share companies are introducing fleets of thesevehicles into urban areas, allowing participants to rentthe scooters for short periods of time. Dockless ridesharing allows the scooters to be left at a finaldestination of the user, ultimately to be retrieved bythe next user or picked up for charging.

Dockless shared e-scooters are touted as a solutionto the last-mile problem, a means to reduce traffic

congestion, and an environmentally preferable modeof transportation [1, 2].While these e-scooters have notailpipe emissions, full consideration of the life cycleimpacts is required to properly understand theirenvironmental impacts. In this study, we use life cycleassessment (LCA) to quantify the total global warm-ing, acidification, eutrophication, and respiratoryimpacts of shared dockless electric scooters. The goalof this study is to identify the key drivers for adverseenvironmental impacts, to offer recommendations onpolicies or practices that would reduce these impacts,and to compare the overall impacts to other modes oftransportation.

OPEN ACCESS

RECEIVED

1May 2019

REVISED

26 June 2019

ACCEPTED FOR PUBLICATION

28 June 2019

PUBLISHED

2August 2019

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2019TheAuthor(s). Published by IOPPublishing Ltd

To the best of our knowledge, this is the first peerreviewed study that comprehensively examines theenvironmental life cycle impacts of shared e-scooters.Life cycle approaches have been used extensively toaddress comparable questions for other transporta-tion technologies. For example, other studies haveexamined alternative transportation options, includ-ing electric vehicles [3, 4], car-sharing programs [5],autonomous vehicles [6], and electric bicycles [7, 8], aswell as full urban transportation systems [9] and com-parisons across two-wheel vehicles [10, 11].

Although it is not peer reviewed, Chester pub-lished the results of an LCA on shared docklesse-scooters that is most relevant to our analysis [12].This study included impacts frommaterials and man-ufacturing, collection and distribution, charging, anddisposal of e-scooters. Results show that manufactur-ing and materials are responsible for the most life-cycle CO2 emissions, followed by collection and dis-tribution and charging of the scooter, for a total of320 g CO2/mile in the best-case scenario. While simi-lar to our study’s goal, our analysis extends this workby collecting detailed primary data for the materialsinventory, daily e-scooter usage, and survey data ontransportation modes being displaced, among otherdifferences.

Compared to e-scooters, far more research hasbeen conducted on the environmental impacts ofbicycles, electric bicycles, mopeds, and motorcycles.In 2001, Zhang et al quantified the life cycle environ-mental impacts of electric bike applications in Shang-hai and found that electrification improved theenvironmental performance of many, but not all,impact categories relative to gasoline-powered motor-bikes [7]. Cherry studied life cycle environmentalimpacts of personal electric two-wheelers (mopedscooters) on the transportation system in China [8],finding that the materials burdens far outweighed theimpacts from assembly, that there were considerableemissions attributable to charging, and that theadverse impacts from the lead batteries was high. (Formodern e-scooters, lead batteries have been replacedwith lithium-ion batteries.) These e-bikes were foundto have lower life cycle emissions than cars per miletraveled across all pollutants examined, withmost life-cycle impacts not from local pollutants, but incurredduring the production process, introducing environ-mental externalities into other regions [8].

Luo et al used a life cycle approach to compare sta-tion-based and dockless bike sharing programs in theUS [13]. They found that rebalancing (collection anddistribution) of shared bicycles was the main source oflife-cycle emissions in dockless bicycles, while thedocking infrastructure was a major source of impactsin station-based bicycles. Consistent with Cherry [8],Luo et al note that car displacement is themost impor-tant factor in reducing emissions, with at least 34% of

bike sharing trips needed to replace car usage in orderto realize net impact reductions [13]. Similarly, Weisset al, Sheng et al, and Rose studied the impacts of per-sonal electric motorcycles and e-bicycles on theenvironment and transportation system, each notingthe importance of the modes of transportation beingdisplaced [11, 14, 15].

In addition to these studies, Hsieh et al used a sys-tem dynamics approach to examine the air pollutionmitigation potential of seated electric scooters in Tai-wan, but limited the scope to use phase impacts [16].Sheng et al [14] compared electric motorcycles togasoline-powered motorcycles on urban noise, find-ing that electrification can reduce noise pollution.Additionally, Bishop et al examined the use phaseenvironmental performance of seated electric scootersin the United Kingdom, while Leuenberger andFrischknect conducted a full comparative LCA for twowheeled vehicles including seated electric scoo-ters [10, 17].

Our study differs from previous research by con-ducting a full LCA to address the environmentalimpacts associated with the materials, manufacturing,transportation, charging, and end-of-life for shareddockless standing e-scooters. Conducting an LCA onan emerging technology offers a greater ability toinform policy makers and consumers, as regulationsand market behavior are still developing [18]. Wenderet al [19] describes approaches to early stage LCA,including the use of scenario development (e.g. [20])to explore the range of possible outcomes. Althoughe-scooter technology is not in an early stage of devel-opment, the business model of shared, docklesse-scooters has emerged quickly in 2018 and is stillevolving. Our study utilizes scenario analysis to betterunderstand the ability to reduce the environmentalimpacts based on this model of shared docklesse-scooters. In this study, we collect primary data one-scooter materials and use, coupled with scenariosand a Monte Carlo analysis to explore the magnitudeand drivers for environmental impacts. In section 2,we describe our data sources and methods. Insections 3 and 4, we present our results and discussion.

2.Methods

In accordance with the ISO standards, our LCAincludes goal and scope, inventory analysis, impactassessment, and interpretation. Figure 1 shows oursystem boundary diagram, which includes materials,manufacturing, e-scooter transport, use and charging,and end-of-life. Due to a lack of data availability andshort scooter lifetime, we exclude routine mainte-nance such as replacing tires or parts during thelifetime of the scooters.

The functional unit for our study is one passenger-mile traveled. Our Base Case assumptions for daily

2

Environ. Res. Lett. 14 (2019) 084031

scooter usage and collection requirements are con-sistent with assumptions for Raleigh, North Carolina.We test a range of parameter values to assess eachinput’s sensitivity and ensure broad applicability ofresults. Equation (1) describes the generalized calcul-ation for impact per passenger-mile for each impactcategory

I represents the life cycle impacts for a given impactcategory (kg-eq/passenger-mile). M represents theburdens associated with the materials and manufac-turing of the scooter (kg-eq/scooter) and T is theburden associated with transportation of the scooterfrom shipping and trucking (kg-eq/scooter). MPSd isthe auto-miles traveled per day (d) for collection anddistribution of scooters (auto-miles/scooter-day) andEFauto is the emissions factor for the vehicle used tocollect the scooters (kg-eq/auto-mile). Egrid,i,d is theelectricity used for charging in hour i, day d (MWh/scooter). EFgrid,i,d represents the emission factor asso-ciated with the specific grid region where the scooter isbeing charged (kg-eq/MWh). Dd represents the

scooter distance traveled on day d. We use TRACI v2.1 characterization factors to convert inventoryresults to environmental impacts [21].

2.1.Materials andmanufacturingTo create an accurate materials inventory for anelectric scooter, we disassembled a Xiaomi M365

scooter, representative of the model that sharedscooter companies including Bird and Lyft currentlydeploy [22]. Table S1 is available online at stacks.iop.org/ERL/14/084031/mmedia in the supportinginformation provides the data for the materialsinventory and the ecoinvent v3.3 process. The materi-als characterization was informed by the documenta-tion provided by the manufacturer and the materialcodes imprinted directly on the components. Themass of each component was recorded to the nearestgram. The ecoinvent material for production ofaluminum alloy was used for the frame, representingthe best ecoinvent material for ‘aerospace grade’aluminum alloy, as described by the manufacturer.

Figure 1. System boundary diagram for a life cycle assessment on shared dockless electric scooters.

IM T MPS EF E EF

D1

dd auto

d igrid i d grid i d

dd

0 0 0 , , , ,

0

å å åå

=+ + ´ + ´( ) ( )

( )

3

Environ. Res. Lett. 14 (2019) 084031

The major materials and components of the e-scooterinclude an aluminum frame (6.0 kg), steel parts(1.4 kg), a lithium ion battery (1.2 kg), an electricmotor (1.2 kg), and tires with tubing (0.83 kg), whichin total account for 89% of the total scooter mass.The lithium-ion battery has a cathode materialLiNi1/3Mn1/3Co1/3O2 (NMC 111), as indicated by thebattery manufacturer. We use the methods detailed byEllingsen et al [23] and Ciez and Whitacre [24] todetermine the environmental impacts of batteryproduction and recycling. Manufacturing burdens areestimated from the ecoinvent process electric bicycleproduction, which is used as a proxy for the energyrequirements to manufacture and assemble the scoo-ter from components. We use a recycled contentapproach; our Base Case assumes 24% recycledcontent for aluminum, consistent withChinese alumi-num in 2017 [25].

2.2. Transportation toUnited StatesWe assume the scooter and battery are assembled inShenzen, China, as indicated by the manufacturer[26]. The total mass of the scooter, packaging, andaccessories is 17.5 kg. We calculate the transportationburdens based on freight shipping to Los Angeles,California (estimated at 11 800 km) and trucking fromLos Angeles to Raleigh, North Carolina (estimated at4000 km), resulting in 207 ton-km and 70 ton-km forshipping and trucking, respectively, per scooter.

2.3. Use phaseThe use phase impacts are influenced by the dailydistance traveled on each e-scooter, the method ofscooter pick-up for charging, the frequency of char-ging, and the time of day and location of charging. Theelectricity impacts of charging use seasonal marginalemissions factors fromAzevedo et al [27], at an eGRIDspatial resolution, which employs a statistical relation-ship between power plant emissions and the hourlygeneration from fossil fuel generators for a region[28, 29]. We assume a charging rate of 84W and a fullbattery charge of 0.335 kWh based on the manufac-turer’s specifications.

E-scooter employees (chargers) can collect anyscooters at any location in the city once the scootersbecome available for collection, without specified col-lection routes, areas for pick-up, or specified scooters.Matching current policy in Raleigh, we assume thatthe e-scooters are picked up each evening to becharged, regardless of the batteries’ state of charge. Todetermine the distribution of the battery state ofcharge at pick-up, we collected end of day (8 pm–

10 pm) data on 800 scooters through the Bird riderapplication.We found that 4.6% of scooters were fullycharged (i.e. unused that day), as shown infigure S1.

We assume that the distribution of personal vehi-cles inWake County, North Carolina is representativeof the vehicles used for collection and distribution.

Using the EPA Motor Vehicle Emission Simulator(MOVES) version 2014a, we determine a global warm-ing (kg CO2-eq/mile) distribution of passenger carand truck emissions for both gasoline and diesel vehi-cles. For respiratory effects (kg PM2.5-eq/mile), acid-ification (kg SO2-eq/mile), and eutrophication (kgN-eq/mile), we use a lognormal distribution withsmall passenger vehicle (EURO4) in ecoinvent 3.3. Tobound the parameters in our analysis, we collecteddata from several employees of shared scooter compa-nies on how many scooters are picked up per trip andthe distance traveled for collection and distribution ofscooters, finding a range of 0.6–2.5 miles per scooterfor collection and distribution.

Given that shared dockless e-scooters are a recentphenomenon, comprehensive data do not yet exist forthe distribution of lifetimes for these products underthese usage conditions. In our analysis, we test a widerange of plausible scooter lifetimes (0.5–2 years),informed by battery lifetimes, the manufacturer war-ranty, and reports of damage under shared usage pro-grams [26, 30]. A 500 cycle lifespan for NMC 111batteries, as specified by the manufacturer, wouldresult in a scooter lifetime of 18 months under a high-usage approach. For the sale of these scooters to indivi-duals, the manufacturer provides a warranty of 12months on the main body and 6 months for the acces-sories [26]. Shared e-scooters may have much shorterlifetimes, however, due to mistreatment or scootersmay last longer under lower usage scenarios. Recentreports have suggested that many scooters may bedamaged by e-scooter users or citizens, and recentreports suggest that e-scooters may have far shorterlifetimes [30].

To better understand the net impacts of e-scooterusage, we compare our results to alternative modes oftransportation. To properly bound this comparison,we conducted a survey of 61 riders and use anotherpublished survey [31] to gain insights into themode oftransportation that e-scooters are replacing (e.g. walk-ing, personal automobile). See tables S6 and S7.

2.4.MonteCarlo analysis and scenariosTo investigate the inherent variability and uncertaintyof several of the parameters used in this study, weconduct a Monte Carlo analysis with assumed dis-tributions for relevant parameters to determine theoverall distribution of life cycle impacts, as shown intable 1. Then, using the Base Case assumptions, we testthe sensitivity of the results to each parameter inisolation to determine which parameters have thegreatest impact on the results.

In addition to the Base Case, we examine three sce-narios relating to the e-scooter collection for chargingand one additional scenario related to e-scooter life-time. In ‘Low Collection Distance,’we assume that theretrieval and distribution distance of e-scooters isreduced, resulting in 0.6 miles driven per scooter by

4

Environ. Res. Lett. 14 (2019) 084031

Table 1. Inputs forMonte Carlo simulations.

Parameters for each impact category Range or (scale, shape factors) Distribution Reference/method

Materials andmanufacturinga (kgCO2-eq/scooter) (200, 17.7) lognormal

(kg PM2.5-eq/scooter) (0.303, 0.0276) lognormal Scooter disassembly, Ecoinvent 3.3

(kg SO2-eq/scooter) (1.74, 0.107) lognormal

(kgNeq/scooter) (1.31, 0.053) lognormal

Collection and distribution emissionsb (kgCO2-eq/mile) (0.366, 0.798) lognormal EPAMOVES 2014bmodel

(kg PM2.5-eq/mile) (2.45× 10−4, 1.21× 10−5) lognormal Ecoinvent 3.3

(kg SO2-eq/mile) (1.33× 10−3, 6.55× 10−5) lognormal Ecoinvent 3.3

(kgN-eq/mile) (8.21× 10−4, 4.05×10−5) lognormal Ecoinvent 3.3

Transportation to theUSc (kgCO2-eq/scooter) (9.54, 3.44) lognormal

(kg PM2.5-eq/scooter) (0.00893, 0.00256) lognormal Distance to ship and truck to Raleigh, Ecoinvent 3.3.

(kg SO2-eq/scooter) (0.0980, 0.0290) lognormal

(kgN-eq/scooter) (0.0130, 0.00397) lognormal

Scooter use andmaintenancec Collection and distribution distance (miles/scooter) 0.6–2.5 uniform Survey of Raleigh chargers

End of day battery charge (%) (0.66, 0.18) lognormal Data collection fromBird application

Scooter lifetime (years) 0.5–2.0 uniform LG specifications forNMCBattery, [20]Time to begin charging (h) 21:00–23:00 uniform Survey, collection schedule

Static variable Value Reference

Scooter distance potential (miles) 18

Energy per full charge (kWh) 0.335 Battery specific values are drawn fromXiaomi’s specifications for them365

scooter

Time to fully charge (h) 4

Power required to charge (kW) 0.08375

a Manufacturing burdens represent a range of 0% to 48% recycled content of aluminum for e-scooter production.

.b Tailpipe emission distribution for CO2-eq/mile are drawn fromMOVES 2014a and modeled in Wake County, NC for passenger cars and trucks. Other impact categories include a distribution range from Ecoinvent passenger vehicles

(EURO4).c Emissions associatedwith transport to theUS are drawn fromEcoinvent processes transport freight by sea and transporting freight by trucking.d 4.6%of scooters are picked upwith a full state of charge. The remaining scooters were fit to a lognormal distribution. See SIfigure S1.

5

Environ.R

es.Lett.14(2019)084031

the collector. In ‘Battery Depletion Limit,’ e-scootersare only retrieved and charged when their battery’sstate of charge drops below 50%. In ‘HighVehicle Effi-ciency,’ the vehicles used for collection have a fuel effi-ciency of 35 miles per gallon (i.e. 235 g CO2-eq/mile).In ‘High Scooter Life,’ we assume that the scooters’lifetime isfixed at two years.

3. Results

Figure 2 shows the life cycle environmental impactsper passenger-mile traveled for each scenario. In theBase Case, the average global warming impact is 202 gCO2-eq/passenger-mile, with 50% from materialsand manufacturing and 43% of impacts coming fromcollection and distribution. The burdens from theelectricity used to charge the scooter contribute only4.7% of the total, while the transportation from themanufacturer proves to be trivial. The error bars infigure 2 represent the range inwhich 95%of theMonteCarlo results fall.

As shown in figures 2(b)–(d), respiratory effects,acidification, and eutrophication are also driven by acombination of the e-scooter materials and manu-facturing and daily collection of the scooters. Using therecycled content approach with 24% recycled content

of aluminum, the aluminum frame and lithium-ionbattery make up 53%–73% of impacts in manufactur-ing and materials across all impact categories. The alu-minum frame is found to be the highest impact driverof respiratory effects, accounting for 46% of thePM2.5-eq from materials and manufacturing, and thebattery pack is found to be the highest driver of acid-ification, accounting for 46% of SO2-eq. Given that thee-scooters are manufactured in China and much of theprimary materials are not sourced from the UnitedStates, these environmental harms are consequentlynot borne by the endusers’ community inour study.

Alternative approaches to collect and distributee-scooters can greatly reduce the adverse environ-mental impacts. Reducing the average driving distancefor collection and distribution to 0.6 miles per scooterreduces the average life cycle globalwarming impacts by27%,while the exclusive use of fuel-efficient vehicles forcollection results in a 12% reduction. Limiting scooterscollection to those with a low battery state of chargewould require a change in policy to allow scooters toremain in public spaces overnight, but could yield a netreduction in globalwarming impacts of 19%.

Figure 3 shows the distribution of the results forscenarios that may reduce global warming impacts. Inall scenarios except High Scooter Life, we observe awide range of outcomes which are driven primarily by

Figure 2. Life cycle environmental impacts for shared electric scooters under Base Case and alternative collection scenarios for(a) global warming, (b) respiratory effects, (c) acidification, and (d) eutrophication. Error bars represent 95%of theMonte Carlosimulations.

6

Environ. Res. Lett. 14 (2019) 084031

the range of scooter lifetimes. The average values foreach scenario, shown by vertical lines in figure 3, arefurther right than the mode value of each scenario dueto short scooter lifetimes which yield a long rightwardtail. Although figure 3 is truncated to more clearly dis-play themean values as vertical lines, the results extendas high as 514 g CO2-eq/passenger-mile. Table S5 inthe supporting information provides the medianvalues for theMonte Carlo results. Due to the long tailof high values for the Base Case, Low Collection

Distance, Battery Depletion Limit, and High VehicleEfficiency scenarios, the median results are 13% to19% lower than the average results. Comparableresults for respiratory impacts, acidification, andeutrophication are shown in the SI, Figures S2–S4.

Figure 4 shows the results of the sensitivity analysison global warming impacts. We see that the globalwarming impacts are most sensitive to the daily usageof the scooter, scooter lifetime, distance driven for col-lection, and vehicle fuel efficiency. Both low daily

Figure 3. Shared e-scooterMonte Carlo analysis of global warming impact. Kernal density functions show the Base Case and fouralternative collection and distribution scenarios: low collection distance, battery depletion limit, high vehicle efficiency, and highscooter lifetime. Colored vertical lines indicate themean value for each scenario.

Figure 4. Sensitivity analysis of global warming potential for various shared e-scooter scenarios. This sensitivity uses the average valuefrom the base case scenario as seen previously infigure 3, but applies a static low and high value to the input parameter of interestshown in parentheses along the y axis.

7

Environ. Res. Lett. 14 (2019) 084031

usage of the scooter and low scooter lifetimes showvery high global warming impacts driven from themanufacturing and materials burdens, which arespread across a smaller number of passenger-milestraveled over the e-scooter lifetime. Figure 4 alsoshows that the results are insensitive to the distance fortransporting the scooter from the manufacturer to thepoint of use and the grid emissions.

While this study was conducted with parametersspecific to Raleigh, North Carolina, the results can beinterpreted and used for a wide range of locations. Wefound that the environmental impacts of the transpor-tation of the scooter from themanufacturer to the enduse location is trivial and the potential differences ingrid emissions for charging the e-scooter yield smallchanges in the overall results. Relative to emissionsfrom charging in Raleigh, charging with a 0 kgCO2/kWh power source (to approximate wind, solar,or nuclear) would decrease life cycle emissions by 6%,while charging with a 1 kg CO2/kWhpower source (toapproximate coal generation)would increase life cycleemissions by 4%. The most important parameter thatwould vary across locations is the collection miles dri-ven per scooter mile. Densely populated metropolitanareas may enable higher densities of e-scooters andlower collection driving distances per scooter. Con-versely, sparsely populated or sprawling areas wouldlikely necessitate higher collection miles driven. Oursensitivity analysis shows that reduced collection dis-tances of 0.6miles per scooter reduce the life cycle CO2

emissions by 27%, while longer driving distances of2.5 miles per scooter increase life cycle CO2 emissionsby 27%.

To better understand the net impacts of sharede-scooter use, we consider the modes of transporta-tion that are being displaced. In our survey ofe-scooter riders, 7% of users reported that they wouldnot have taken the trip otherwise, 49% would havebiked or walked, 34% would have used a personal

automobile or ride-share service, and 11%would havetaken a public bus (table S7). These results are con-sistent with a survey conducted in Portland, Oregon,which shows 8% would not have taken the trip, 45%would have biked or walked, 36% would have used anautomobile, and 10%would have used a bus or street-car [31]. To estimate the displaced burdens frome-scooter usage, we assume that each passenger-mileon an e-scooter displaces 0.34 passenger-miles in apersonal car, 0.11 passenger-miles on a public bus, and0.08 miles on a bicycle. We also assume the life cycleglobal warming impacts of personal car use is 414 gCO2-eq/passenger-mile, using Argonne NationalLaboratory’s GREET 2 model with US average petro-leummix, vehiclemodel year 2012, 26miles per gallonefficiency, and one passenger [32].We assume impactsfrom bus ridership is 82 g CO2-eq/passenger-mile[33], consistent with the well-to-wheels calculation forurban diesel bus use during peak hours from Chesterand Horvath, 2009, with the important caveat thatemissions from buses do not decrease proportionallywith the loss of one rider. We assume that the use of apersonal bicycle results in 8 g CO2-eq /passenger-mile[11]. Using these assumptions, we calculate that theavoided life cycle emissions from car and bus use is150 g CO2/passenger-mile, which we term the‘Benchmark Displacement.’ This Benchmark Dis-placement rate is 26% lower than the average BaseCase impacts associated with the use of sharede-scooters and very near the High Scooter Life andLowCollectionDistance scenarios.

In table 2, we present the likelihood that thee-scooter life cycle global warming impacts per pas-senger-mile traveled exceeds the impacts associatedwith the BenchmarkDisplacement and alternative sin-gle modes of transportation. For this assessment, weuse representative life cycle emissions values for thesealternatives and report the share of e-scooter MonteCarlo analysis results that exceed those values.

Table 2. Likelihood that the e-scooter life cycle global warming impacts per passenger-mile traveled exceeds the impacts associatedwithalternativemodes of transportation.

Base case

Low collection

distance

Battery depletion

limit

High vehicle

efficiency

High scooter

lifetime

Personal automobilea (414 gCO2/mi) 1.7% 0.3% 0.7% 1.0% 0.0%

Shared dockless bicycleb (190 gCO2/mi) 33.2% 20.9% 23.6% 30.0% 0.0%

BenchmarkDisplacementc (150 gCO2/mi) 65.0% 34.8% 39.9% 50.0% 4.0%

Electricmopedd (119 gCO2/mi) 100.0% 54.2% 66.9% 89.5% 100.0%

Buswith high ridershipe (82 gCO2/mi) 100.0% 99.6% 100.0% 100.0% 100.0%

Electric bicycled (40 gCO2/mi) 100.0% 100.0% 100.0% 100.0% 100.0%

Bicycled (8 gCO2/mi) 100.0% 100.0% 100.0% 100.0% 100.0%

Values drawn froma ArgonneNational LabGREET 2model.b Reference [13].c Benchmark Displacement assumes 1 e-scooter passenger-mile displaces 0.34 miles of personal automobile travel, 0.11 miles of bus travel,

and 0.08miles of bicycle travel.d Reference [11].e Reference [12].

8

Environ. Res. Lett. 14 (2019) 084031

The personal automobile, bus with high ridership,and bicycle emissionsmatch those previously describedin the calculation of the Benchmark Displacement. Theshared dockless bicycles represent non-electric bikesthat require ‘rebalancing’ [13]. The electric moped,electric bicycle, and bicycle values represent personalownership,whichdonot require rebalancing [11].

These results show that dockless e-scooters con-sistently result in higher life cycle global warmingimpacts relative to the use of a bus with high ridership,an electric bicycle, or a bicycle per passenger-mile tra-veled. However, choosing an e-scooter over driving apersonal automobile with a fuel efficiency of 26 milesper gallon results in a near universal decrease in globalwarming impacts. The use of dockless e-scooters areoften preferable to dockless bicycles, yielding lower lifecycle emissions 67% to 100% of the time across thescenarios. When compared to the Benchmark Dis-placement CO2 emissions, our Base Case shows a 65%chance that the life cycle e-scooter emissions will behigher. This likelihood is reduced, but nontrivial, forour LowCollection Distance (35%), Battery DepletionLimit (40%), High Vehicle Efficiency (50%), and HighScooter Lifetime (4%) scenarios. These results under-score the importance of ensuring long lifetimes fore-scooters in reducing life cycle emissions.

4.Discussion

In this study, we found that the global warmingimpacts associated with the use of shared e-scootersare dominated by materials, manufacturing, andautomotive use for e-scooter collection for charging.Increasing scooter lifetimes, reducing collection anddistribution distance, using more efficient vehicles,and less frequent charging strategies can reduceadverse environmental impacts significantly. Withoutthese efforts, our Base Case calculations for life cycleemissions show a net increase in global warmingimpact when compared to the transportationmethodsoffset in 65% of our simulations. Taken as a whole,these results suggest that, while e-scooters may be aneffective solution to urban congestion and last-mileproblem, they do not necessarily reduce environmen-tal impacts from the transportation system.

Cities that seek to integrate e-scooters into theirtransportation system have several policy optionsavailable to reduce the life cycle environmental bur-dens associated with their use. Allowing e-scooters toremain in public areas overnight would decrease theautomobile burdens associated with picking up fullycharged or nearly fully charged e-scooters. Requiringcentral management or improved e-scooter collectionprocesses could reduce the auto-miles traveled for col-lection and distribution. Additionally, cities couldenact or enforce anti-vandalism policies to reducee-scooter misuse or mistreatment which can result in

short lifetimes (and thus high materials and manu-facturing burdens per passenger-mile traveled).

The scooter companies also can take meaningfulaction to reduce the life cycle burdens of their pro-ducts. They can reduce collection and distributionburdens by incentivizing or requiring the use of effi-cient automobiles. In addition, they could reducevehicle miles traveled for collection and distributionthrough centralized management or by allowing char-gers to ‘claim’ e-scooters to eliminate unnecessary andcompetitive driving during daily collection.

This study clearly demonstrates that there is thepotential for e-scooters to increase life cycle emissionsrelative to the transportationmodes that they displace.Although we use a Monte Carlo analysis withinformed ranges for input parameters such as scooterlifetime, collection distance, and vehicle efficiency,cities and e-scooter companies alike can use this studyto further explore life cycle impacts of e-scooters witha higher level of detail in the future. Claims of environ-mental benefits from their use should be met withskepticism unless longer product lifetimes, reducedmaterials burdens, and reduced e-scooter collectionand distribution impacts are achieved.

Acknowledgments

The authors would like to thank Dr AndrewGrieshop,Dr Joseph DeCarolis, Dr Jarod Kelly, and Aditya Sinhafor their valuable insights. This work was conductedwith financial support from the North Carolina StateUniversity Department of Civil, Construction, andEnvironmental Engineering Research Experience forUndergraduates Program.

References

[1] SipeN and Pojani D 2018Can E-Scooters Solve the ‘LastMile’Problem? They’ll Need toAvoid the Fate of Dockless Bikes.TheConversation September 2018. (https://theconversation.com/can-e-scooters-solve-the-last-mile-problem-theyll-need-to-avoid-the-fate-of-dockless-bikes-102633)

[2] SmithC S and JosephP S 2018 E-scooter scenarios: evaluatingthe potentialmobility benefits of shared dockless scooters inChicagoChaddick Institute Policy Series (Chicago, IL: DePaulUniversity)

[3] Hawkins TR et al 2013Comparative environmental life cycleassessment of conventional and electric vehicles J. Ind. Ecol. 1753–64

[4] Michalek J J et al 2011Valuation of plug-in vehicle life-cycle airemissions and oil displacement benefitsProc. Natl Acad. Sci.108 16554–8

[5] ChenTD andKockelmanKM2016Carsharing’s life-cycleimpacts on energy use and greenhouse gas emissionsTransp.Res.D 47 276–84

[6] Gawron JH,KeoleianGA,DeKleine RD,Wallington T J andKimHC2018 Life cycle assessment of connected andautomated vehicles: sensing and computing subsystem andvehicle level effects Environ. Sci. Technol. 52 3249–56

[7] ZhangC,WangC, Sullivan J, HanWand SchuetzleD 2001 Lifecycle assessment of electric bike application in Shanghai SAETechnical Paper 2001-01-3727 SAE International (https://doi.org/10.4271/2001-01-3727)

9

Environ. Res. Lett. 14 (2019) 084031

[8] Cherry CR 2007 Electric two-wheelers inChina: analysis ofenvironmental, safety, andmobility impactsDiss. Institute ofTransportation Studies (Berkeley, CA :University of CaliforniaPress )

[9] ChesterMV,HorvathA andMadanat S 2010Comparison oflife-cycle energy and emissions footprints of passengertransportation inmetropolitan regionsAtmos. Environ. 441071–9

[10] LeuenbergerM and Frischknecht R 2010 Life Cycle Assessmentof TwoWheel Vehicles (Uster: ESU-Service Ltd )

[11] WeissM,Dekker P,MoroA, ScholzH and PatelMK2015Onthe electrification of road transportation–a review of theenvironmental, economic, and social performance of electrictwo-wheelersTransp. Res.D 41 348–66

[12] ChesterM2019 It is a BirdKIt is a LimeKIt is DocklessScooters! But CanThese Electric-PoweredMobilityOptionsBeConsidered SustainableUsing Life-Cycle Analysis?(https://chesterenergyandpolicy.com/2019/01/28/its-a-bird-its-a-lime-its-dockless-scooters-but-can-these-electric-powered-mobility-options-be-considered-sustainable-using-life-cycle-analysis/)

[13] LuoH,KouZ, Zhao F andCaiH 2019Comparative life cycleassessment of station-based and dock-less bike sharing systemsRes. Conservation Recycling 146 180–9

[14] ShengN, ZhouX andZhouY 2016 Environmental impact ofelectricmotorcycles: evidence from traffic noise assessment bya building-based datamining technique Sci. Total Environ. 55473–82

[15] RoseG2012 E-bikes and urban transportation: emergingissues and unresolved questionsTransportation 39 81–96

[16] Hsieh PY, Chang L FW, YuTY andWuKC2018 Evaluationofmitigation effects on air pollutants for electric scooters inTaiwanwith the energyflow analysis and systemdynamicsapproach IOPConf. Ser.: Earth Environ. Sci. 191 012136

[17] Bishop JD,Doucette RT, RobinsonD,Mills B andMcCullochMD2011 Investigating the technical, economicand environmental performance of electric vehicles in the real-world: a case study using electric scooters J. Power Sources 19610094–104

[18] Miller S A andKeoleianGA2015 Framework for analyzingtransformative technologies in life cycle assessment Environ.Sci. Technol. 49 3067–75

[19] Wender BA, Foley RW,Hottle TA, Sadowski J,Prado-LopezV, EisenbergDA, Laurin L and Seager TP 2014Anticipatory life-cycle assessment for responsible research andinnovation J. Responsible Innov. 1 200–7

[20] EckelmanM J,MauterMS, Isaacs J A and ElimelechM2012Newperspectives on nanomaterial aquatic ecotoxicity:production impacts exceed direct exposure impacts for carbonnanotoubesEnviron. Sci. Technol. 46 2902–10

[21] Bare J C 2012Tool for the Reduction andAssessment ofChemical andOther Environmental Impacts (TRACI),Version 2.1 -User’sManual; EPA/600/R-12/554 2012

[22] Whittaker Z 2018XiaomiDoes notWant Lyft Using ItsElectric Scooters. TechCrunch.October 2018 (https://techcrunch.com/2018/10/31/xiaomi-doesnt-want-lyft-using-its-electric-scooters/)

[23] Ellingsen LAW,Majeau‐Bettez G, Singh B, Srivastava AK,Valøen LO and StrømmanAH2014 Life cycle assessment of alithium‐ion battery vehicle pack J. Ind. Ecol. 18 113–24

[24] Ciez R E andWhitacre J F 2019 Examining different recyclingprocesses for lithium-ion batteriesNat. Sustainability 2 148

[25] BertramM,Ramkumar S, RechbergerH, RombachG,Bayliss C,MartchekK J,Müller DB and LiuG 2017Aregionally-linked, dynamicmaterialflowmodelling tool forrolled, extruded and cast aluminiumproductsRes.Conservation Recycling 125 48–69

[26] XiaomiUnited States 2019 (https://mi.com/us/mi-electric-scooter/specs/) (Accessed: 02 July 2019)

[27] Azevedo I L,HornerNC, Siler-Evans K andVaishnav PT 2017ElectricityMarginal Factor Estimates. Center For Climate andEnergyDecisionMaking (Pittsburgh: CarnegieMellonUniversity)

[28] Siler-Evans K, Azevedo I L andMorganMG2012Marginalemissions factors for theUS electricity system Environ. Sci.Technol. 46 4742–8

[29] Siler-Evans K, Azevedo I L,MorganMGandApt J 2013Regional variations in the health, environmental, and climatebenefits fromwind and solar generation Proc. Natl Acad. Sci.110 11768–73

[30] Newberry L 2018 Fed-up Locals Are Setting Electric Scooterson Fire and Burying Them at Sea. Los Angeles Times, August10, 2018.

[31] Portland Bureau of Transportation 2018 2018 E-Scooter PilotUser Survey Results (https://www.portlandoregon.gov/transportation/article/700916)

[32] ArgonneNational Laboratory, TheGreenhouseGases,Regulated Emissions, and EnergyUse in TransportationModel(GREET), GREET2, 2018

[33] ChesterMVandHorvath A 2009 Environmental assessmentof passenger transportation should include infrastructure andsupply chainsEnviron. Res. Lett. 4 024008

10

Environ. Res. Lett. 14 (2019) 084031