Examining the Impact of Weather, Temporal Factors, and User

Traits on Multimodal Shared Micromobility Systems in Non-Urban

Campus Environments: The MORE Sharing Case Study

Maryna Pobudzei

and Silja Hoffmann

Professorship for Intelligent, Multimodal Transportation Systems, University of the Bundeswehr Munich,

Werner-Heisenberg-Weg 39, 85577 Neubiberg, Germany

Keywords: Shared Micromobility, non-Urban Contexts, University Campus, Corporate Campus, Neighborhoods,

Communities, E-Bikes, E-Cargo Bikes, E-Scooters, E-Mopeds, Machine Learning Models, Weather Impact,

Time Impact, More Sharing.

Abstract: Although shared micromobility systems in cities have been extensively studied, their potential for non-urban

settings such as university campuses and rural communities has not been explored much yet. This study aims

to fill this gap by examining a multimodal shared micromobility service that offers various options through a

single app, such as city bikes, e-bikes, e-cargo bikes, e-mopeds, and e-scooters. The study analyzed this

campus-based system's first four months, considering factors like weather, time, user demographics, pre-

reservation duration, and vehicle types. Machine learning models like Negative Binomial Regression,

Random Forests, Gradient Boosted Regression Trees, and Neural Networks were used to analyze the data.

The study found that e-scooters were the most popular, followed by e-bikes. E-mopeds were used less but

were reserved for more extended periods. Most trips were taken on weekdays, especially between 8 AM and

6 PM. Reservation numbers peaked in the first month, and subsequent months showed longer reservation

durations and distances. Rain decreased trip numbers and distances but increased reservation durations.

Reservations on Fridays, weekends, and holidays were shorter but covered more distance. Female users

tended to travel longer distances. These findings can benefit similar non-urban environments, broadening the

application of shared micromobility systems.

1 INTRODUCTION

The 21st century has seen a rise in shared

micromobility as a form of urban transportation.

While most research has focused on city-wide

commercial projects facilitated by private mobility

companies, there is potential for implementing shared

micromobility solutions in smaller settings such as

residential neighborhoods, non-urban communities,

campuses, and corporate settings. These unique

contexts present opportunities and challenges that

require thorough investigation, as successful

implementation requires a deep understanding of

local dynamics and community requirements.

The University of the Bundeswehr in Munich,

Germany, is a suitable case study for exploring shared

micromobility in microenvironments. This 140-

https://orcid.org/0000-0002-3219-9144

https://orcid.org/0000-0002-0499-0342

hectare campus is one of the largest campus-based

institutions in Germany, with a diverse population of

5,300 members. This population comprises 72%

students, 16% academic staff, 8% non-academic

personnel, and 4% professors (UniBw, 2023b). The

student body comprises military officers and civilians

from various regions and countries, primarily aged

20-30, within the middle-income bracket, and

predominantly residing on campus. Approximately

27% of students and 25% of academic staff are female

(UniBw, 2023a).

To meet the transportation needs of this

population, the university recently introduced

"MORE Sharing," a micromobility sharing system

operated by a third-party contractor (evhcle, 2023;

Huber, 2023; MORE Sharing, 2023; Pobudzei,

Kemmerzehl, et al., 2023). Users can book and access

a variety of vehicles via a mobile app, which displays

Pobudzei, M. and Hoffmann, S.

Examining the Impact of Weather, Temporal Factors, and User Traits on Multimodal Shared Micromobility Systems in Non-Urban Campus Environments: The MORE Sharing Case Study.

DOI: 10.5220/0012341900003702

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 10th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2024), pages 193-203

ISBN: 978-989-758-703-0; ISSN: 2184-495X

193

rental prices and automatically selects the most cost-

effective rate based on the rental duration. Users are

also provided with the option of a complimentary

monthly mobility budget after registration and

payment details submission. Additional credits can be

earned by participating in surveys offered by the

service (evhcle, 2023; Pobudzei, Wichmann, et al.,

2023).

MORE Sharing users are not confined to the

operating zone during their rides and can venture

beyond it, with the app offering a park mode for

breaks without terminating the rental. However,

rentals must be concluded within the operating zone,

requiring users to ensure sufficient battery charge for

their return trip (evhcle, 2023). This integrated

multimodal micromobility system, coupled with the

unique offering of a free mobility budget, sets MORE

Sharing apart from other platforms and represents an

area of shared micromobility that has received limited

exploration (Pobudzei, Wichmann, et al., 2023).

This study aims to provide insights into the

utilization patterns and influential factors of shared

micromobility systems, thereby contributing to the

development of sustainable transportation initiatives

in microenvironments like campuses, corporate

landscapes, residential neighborhoods, and non-urban

communities. Specifically, this paper evaluates

metrics such as hourly trip counts, trip durations, and

reservation lengths during the initial months of the

micromobility sharing system's operation on the

university campus.

The analysis considers variables such as weather

conditions, time of day, vehicle type, reservation

duration, and user demographics, considering their

potential impact on these metrics. Advanced

analytical techniques, including Negative Binomial

Regression (NBR), Random Forests (RF), Gradient

Boosted Regression Trees (GBRTs), and Neural

Networks (NN), are employed for this analysis. The

outcomes of this research provide valuable insights

for policymakers, urban planners, and transportation

providers, enhancing shared micromobility system

design and implementation across various settings,

including residential areas and large campus

environments.

2 LITERATURE REVIEW

Shared micromobility systems are becoming

increasingly common in urban areas. However, these

systems are typically single-mode, meaning that each

service provider's app only offers a specific mode of

transport, such as city bikes, e-bikes, e-scooters, e-

cargo bikes, or e-mopeds. Users must register with

multiple providers to access different mobility

options.

Users' demographics for each shared mode of

transport vary depending on the provider and location

(Pobudzei, Wichmann, et al., 2023). For example,

shared city bikes, e-bikes, and e-scooters are more

popular among younger adults and men, while

women and older populations use them less

frequently (NACTO, 2022; Rérat, 2021).

In the United States, e-scooter users and bike

share members typically embark on rides lasting 11-

15 minutes, covering up to 3 kilometers (NACTO,

2022; Younes et al., 2020). Station-based bike share

users usually opt for longer trips, lasting 24-28

minutes and covering approximately 5 kilometers

(NACTO, 2022). However, data on usage patterns for

shared e-cargo bikes and e-mopeds is limited,

indicating a gap in current research (Pobudzei,

Kemmerzehl, et al., 2023).

In Munich, shared e-scooters are most frequently

used on Friday and Saturday afternoons, with longer

trips taken on weekends and holidays than on

workdays (Pobudzei et al., 2022; Schreier et al., 2022;

Tießler et al., 2023). The use of shared e-scooters

witnesses an increase in frequency and duration from

July to November, in contrast to the winter months

(Pobudzei et al., 2022).

Shared city bikes and e-bikes see higher usage

rates on weekdays, especially during peak commute

hours (Fishman, Washington, & Haworth, 2015;

Fishman, Washington, Haworth, et al., 2015; Rérat,

2021; Younes et al., 2020). However, some cities also

witness a surge in usage during lunchtime and

weekends, catering to recreational purposes

(Pobudzei, Wichmann, et al., 2023). Shared e-cargo

bikes are typically used on weekdays for commercial

and delivery purposes (Becker & Rudolf, 2018),

while shared e-mopeds e-mopeds are popular for

weekday commuting and recreational use on

weekends and evenings (Pobudzei, Wichmann, et al.,

2023).

Weather conditions also play a significant role in

the usage of shared micromobility. Extreme

temperatures and adverse weather conditions like

high winds, rain, snow, and other precipitation

discourage users due to safety and comfort concerns

(Gebhart & Noland, 2014; Noland, 2021; Pobudzei et

al., 2022).

Despite the literature on shared micromobility in

urban settings, more research should be conducted on

systems deployed in non-urban settings and

microenvironments such as university or corporate

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

194

campuses, residential neighborhoods, or non-urban

communities.

This paper addresses this gap by focusing on the

shared micromobility service at the University of the

Bundeswehr in Munich, where users can access

multiple micromobility modes within a single

application. The investigation aims to analyze hourly

trip counts, trip distances, and durations for shared city

bikes, e-bikes, e-cargo bikes, e-scooters, and e-mopeds

within a micromobility sharing system in the campus

environment. The findings of this study could be

helpful in similar environments, thereby expanding the

applicability of shared micromobility systems.

3 DATA AND METHODS

3.1 MORE Sharing Setup and Data

On the 6

of March, 2023, the MORE Sharing service

was officially launched and communicated to the

university community via email notification. Soon

after, on the 15

of March, a launch event was

organized to introduce potential users to the service.

The event assisted with the app installation, a

comprehensive tutorial on how to use it, and an

introduction to the range of available vehicles.

Additionally, a dedicated webpage with instructional

videos and frequently asked questions was established

and made accessible to help users understand the

service (Huber, 2023; MORE Sharing, 2023).

The MORE Sharing service is only available to

members affiliated with the University of the

Bundeswehr, including students and staff. Upon

registration, users can choose a mobility budget of up

to 300 Euros, automatically renewing monthly. The

unused monthly mobility budget cannot be

transferred to the following months. Users can also

earn extra credits by participating in surveys. The

pricing structure for MORE Sharing is 0.13 Euros per

minute for riding and parking any vehicle, with a

daily cap of 25 Euros. The system applies the most

economical rate, minute-based or daily, depending on

the rental duration. Users can reserve a vehicle for up

to 15 minutes at no cost before starting their ride.

From March to June 2023, MORE Sharing

provided 94 vehicles, including 25 city bikes, 24 e-

bikes, 7 e-cargo bikes, 9 e-mopeds, and 29 e-scooters.

Users can start a ride by scanning a vehicle's QR code

or selecting it directly from the map view in the app.

After confirming the rental, the digital lock is

activated. Users can pause and end their rides using

the MORE Sharing app. For city bikes or e-cargo

bikes, users need to manually secure the lock, while

e-bikes, e-scooters, and e-mopeds use automatic

locking mechanisms. People who want to ride e-

mopeds must have a valid driver's license, which must

be verified within the app, and are required to wear a

helmet.

The MORE Sharing service operates within the

university campus and extends to specific public

transport stops within a 3-kilometer radius. Users can

start or end their rides at these stops. However, they

must finish their trips within the designated

operational area, even if they travel beyond the 3-

kilometer radius. Users can activate the parking mode

if they need to park their vehicles outside the

operational area (MORE Sharing, 2023a; Pobudzei,

Kemmerzehl, et al., 2023).

The MORE Sharing app collects reservation

details such as the reservation time, user ID, vehicle

ID, start and end times of each journey, and the initial

and final mileage readings of the vehicles. Between

March and June 2023, we analyzed the data to

calculate the frequency of trips, distance covered, and

rental duration for each vehicle type hourly. Please

note that the MORE Sharing service was not available

on May 17

, 18

, and June 24

, 2023, due to local

on-campus events.

Apart from the reservation data, we also collected

meteorological data from weather station ID 3,379,

located in Munich City at coordinates 48.16 latitude,

11.54 longitude, and an elevation of 515 meters. This

weather dataset included meteorological parameters

such as wind chill index, relative humidity, and

recorded precipitation levels (DWD, 2023).

3.2 Modeling Methods

The study aims to analyze the patterns of hourly trip

counts, trip distances, and durations observed during

the initial four months of operation of a

micromobility sharing system deployed on a

university campus. The study's main objectives are to

gain insights into user behaviors, evaluate the

system's effectiveness, and comprehensively

understand usage patterns.

In addition to descriptive methods, the study

explored the application of various machine learning

models, including Negative Binomial Regression

(NBR), Random Forests (RF), Gradient Boosted

Regression Trees (GBRTs), and Neural Networks

(NN). The models were trained using 80% of the

dataset and tested on the remaining 20%. The random

seed was set to 42 to ensure consistent results.

The NBR model is suitable for counting data that

shows overdispersion, where the variance exceeds the

mean. It can effectively handle categorical and

Examining the Impact of Weather, Temporal Factors, and User Traits on Multimodal Shared Micromobility Systems in Non-Urban Campus

Environments: The MORE Sharing Case Study

195

continuous predictor variables (Noland, 2021;

Pobudzei et al., 2022).

RF constructs numerous decision trees by

leveraging random data subsets and combines the

predictions of individual trees to produce a final

forecast. This approach mitigates overfitting and

enhances the model's robustness by selecting random

feature subsets for each tree. RF demonstrates

proficiency in handling continuous and categorical

variables, allowing it to model complex non-linear

relationships (Breiman, 2001).

GBRTs adopt an iterative approach, initially

fitting a simple decision tree to the data and then

concentrating on the areas where the model exhibits

weaknesses. These problematic instances are

assigned higher weights, prompting the training of

new decision trees to refine the model's predictions.

GBRTs typically excel when dealing with smaller

datasets, although they demand more computational

resources for training compared to other methods

(Davis, 2014).

Lastly, NN leverages a multilayered architecture

comprising numerous interconnected neurons.

During training, the weights that control the strength

and direction of signal transmission between neurons

are adjusted to enhance the model's predictive

capabilities. NNs are proficient in modeling complex

non-linear relationships and can handle diverse input

data types, including images, text, and numerical

data. It is important to note that they come with higher

computational demands, mainly when applied to

larger datasets (Analytics Vidhya, 2023).

4 RESULTS AND DISCUSSION

By June 2023, the number of registered users for

MORE Sharing had grown to 2,379. There was a

significant surge in registrations during the first three

weeks after the service launch (Figure 1). All

registered users received a monthly mobility budget

of 300 euros that was automatically renewed

monthly. The unused monthly mobility budget cannot

be transferred to the following months. Figure 1

shows that the majority of users fall within their 20s,

and 81.8% of users are male, 17.7% are female, and

0.4% identify as diverse, reflecting the demographics

of the university community.

Between March and June 2023, MORE Sharing

facilitated 25,742 distinct trips, with an average of

approximately 238 daily trips. 66% of registered

users utilized the service during this time, and 1,570

unique customers engaged with the service. Table 1

analyzes the cumulative trip count, distance traveled,

and reservation duration for individual customers

during the periods of March–April 2023 and May–

June 2023. Specifically, during March–April 2023,

1,242 unique customers engaged with MORE

Sharing, while the number increased to 1,333

customers during May–June 2023. Although May–

June witnessed an increase in active customers, some

users who had availed the service in March–April did

not do so in May–June, resulting in a cumulative

count of 1,570 customers encompassing both periods.

During the first two months, customers took an

average of approximately 10.8 trips each. However,

in the following period, the average number of trips

per customer decreased to approximately 9.2. In the

first two months of operation, 75% of customers

made 15 trips or fewer. This number decreased to 12

trips or fewer in the following period. Despite this

trend, some customers used the service much more

frequently than the average. These observations

indicate a broad spectrum of user engagement levels,

behaviors, and preferences, ranging from less active

members to prolific users.

Table 1 shows that the cumulative distance

traveled and reservation duration per customer

increased overall from the first to the second period.

This suggests that users optimize their utilization

within each trip, resulting in an overall increase in

distance covered and reservation duration. A subset

of users covered significantly longer distances in both

periods, indicating potential usage of MORE Sharing

for journeys beyond the immediate campus vicinity.

Additionally, some users preferred reserving vehicles

for extended durations, possibly for activities or

events requiring prolonged mobility. Extended

reservations per customer are also aligned with the

fact that MORE Sharing users had a monthly mobility

budget of 300 euros.

Between March and June 2023, MORE Sharing

boasted a fleet comprising 94 vehicles, encompassing

25 city bikes, 24 e-bikes, 7 e-cargo bikes, 9 e-mopeds,

and 29 e-scooters. To gauge the utilization rates for

each vehicle type, we calculated the ratio of daily

trips to the number of available vehicles of that

particular type (Figure 2). These metrics revealed that

e-scooters were the most frequently utilized vehicle

type within the MORE Sharing system, with e-bikes

closely following (Figure 2). This trend, especially

evident in the initial weeks of system operation,

indicates that e-scooters and e-bikes enjoy heightened

popularity among users, presumably due to their

convenience and ease of use. E-cargo bikes and city

bikes exhibited similar utilization patterns, while e-

mopeds recorded the lowest average utilization

(Figure 2).

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

196

Figure 1: MORE Sharing user registrations and

demographics

Table 1: Cumulative sum of trips, driven distance, and

reservation duration per customer.

Parameters Cumulative sum per customer

Parameter Period

Active

customers

Min 25% Mean 50% 75% Max

Trips

3. -

4.2023

1,242 1 3 10.8 7 15 103

5. -

6.2023

1,333 1 2 9.2 5 12 87

Driven

Distance,

Kilometers

3. -

4.2023

1,242 0.0 5 23 13 31 325

5. -

6.2023

1,333 0.0 4 24 11 29 432

Reservation

Duration,

Minutes

3. -

4.2023

1,242 0.2 34 577 134 419 17,308

5. -

6.2023

1,333 0.3 31 861 145 623 25,897

However, the lower average utilization of e-

mopeds does not mean a lack of popularity. Users

often reserved e-mopeds for extended periods

compared to other vehicle types (Table 2). E-cargo

bikes follow suit, with an average reservation

duration approximately three times shorter than e-

mopeds. E-bikes and e-scooters displayed

comparable average reservation durations, while city

bikes featured the shortest mean reservation duration

(Table 2).

An analysis of driven distances, as presented in

Table 2, indicates that e-mopeds typically cover the

greatest mean distance, implying their use for longer

journeys compared to other vehicle types.

Nonetheless, it is crucial to note that half of the trips

made on e-mopeds are less than or equal to 2

kilometers, suggesting a mix of shorter and longer

trips. E-bikes, e-cargo bikes, and e-scooters exhibit

comparable average distances traveled, while city

bikes demonstrate the lowest mean distance traveled.

Figure 3 offers an overview of trip distribution

across different distance and duration categories for

each vehicle type. Predominantly, short trips in both

duration and distance categories dominate the usage

patterns across all vehicle types within MORE

Sharing. This observation underscores the primary

utilization of the service for brief journeys within the

campus vicinity or nearby areas.

Extended reservation durations do not necessarily

correlate with greater distances traveled, as these

often involve extended idle times rather than

continuous movement (Figure 3). This observation

suggests that users reserve vehicles for longer

durations to accommodate their needs, including

extended breaks or multiple stops during their

journeys, rather than solely focusing on covering

longer distances. E-mopeds, e-cargo bikes, and e-

bikes are utilized for longer distances during longer-

duration rentals (Figure 3), indicating that users

prefer these vehicle types for more extensive journeys

or tasks requiring greater distances.

Kernel density estimate (KDE) plots (Figure 4)

depict the distribution of trips across different

micromobility modes throughout the week and at

various times of the day. Most trips, irrespective of

vehicle type, occur on weekdays, in line with the

primary user base consisting of students, faculty, and

staff members who are more active on campus during

weekdays.

Trip numbers rise between 6 AM and 8 AM,

corresponding to the morning commute on campus.

From 8 AM to 6 PM, the trip counts for all

micromobility modes remain consistently high. After

7 PM, the number of trips declines, indicating

reduced demand during the evening hours when users

likely conclude their campus activities. E-mopeds

exhibit a distinct peak in usage between 1 PM and 6

PM, possibly due to various reasons, such as using e-

mopeds for longer trips, leisure activities, or running

errands during the mid-afternoon period.

The study investigated the effects of time,

weather, user demographics, and vehicle types on

MORE Sharing's hourly trip counts, reservation

duration, and driven distance (Table 3). Results from

the Negative Binomial Regression (NBR) model for

hourly trip counts revealed that precipitation has a

negative impact on the number of shared

micromobility trips (Table 3).

Unfavorable weather conditions or safety

concerns may decrease trip counts during wet

conditions (Table 3). However, parameters like wind

chill index and humidity level did not significantly

affect hourly trip rates. Comparing the number of

trips per hour for different vehicle types (Table 3), it

was found that e-scooters, e-bikes, and city bikes had

higher trip counts compared to e-cargo bikes. This

discrepancy could be attributed to factors such as

vehicle availability or specific use cases contributing

to differences in trip counts.

Examining the Impact of Weather, Temporal Factors, and User Traits on Multimodal Shared Micromobility Systems in Non-Urban Campus

Environments: The MORE Sharing Case Study

197

Figure 2: Daily utilization rate per vehicle type.

Table 2: Reservation duration (minutes) and driven distance (kilometers) per vehicle type.

Parameter Vehicle type

Number of

trips

Min 25% Mean 50% 75% Max

Reservation

duration, minutes

City Bike 4,803 0.1 4.3 83.3 8.0 26.8 16,024.7

E-Bike 7,769 0.0 4.3 55.4 8.2 32.2 8,924.7

E-Cargo Bike 1,525 0.1 5.2 95.1 12.2 47.4 4,460.0

E-Moped 977 0.1 2.2 262.1 20.9 107.1 9,796.7

E-Scooter 10,668 0.0 4.4 59.3 8.5 33.5 10,197.8

Trip distance,

kilometers

City Bike 4,803 0.0 0.5 1.6 1.0 1.7 38.5

E-Bike 7,769 0.0 0.8 2.3 1.4 2.6 57.0

E-Cargo Bike 1,525 0.0 0.5 2.3 1.2 2.7 63.5

E-Moped 977 0.0 0.0 5.8 2.0 8.0 58.0

E-Scooter 10,668 0.0 0.9 2.5 1.6 3.0 38.3

Figure 3: Distribution of MORE Sharing trips across different distance and duration categories.

Figure 4: Kernel density plots (mode independent): distribution of MORE Sharing trips.

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

198

Table 3: Modeling hourly trip counts, reservation durations, and trip distances with Negative Binomial Regression (NBR).

Bolded values indicate statistical significance at p < 5%.

Hourly trip counts Reservation duration, minutes Trip distance, kilometers

N 5,325 20,516 20,516

RMSE 2.63 324.7 3.24

Variable coef z coef z coef z

Vehicle pre-reserved, minutes

0.01 6.86 0.04 16.86

User age

0.02 12.94 0.00 0.11

User registered since, days

0.01 13.20 0.00 -3.60

Female

0.04 1.48 0.07 2.49

Humidity, % -0.01 -0.46 0.00 -0.76 0.00 -0.30

Precipitation index -0.19 -4.44 0.17 9.25 -0.10 -4.36

Wind chill index 0.00 0.81 0.01 33.72 0.00 7.08

E-Bike 0.95 17.00 -0.64 -20.37 -0.01 -0.39

City Bike 0.64 11.21 -0.21 -6.38 -0.39 -9.66

E-Moped 0.11 1.40 0.85 18.34 0.85 16.39

E-Scoote

1.32 23.73 -0.53 -17.29 0.07 1.92

3-5AM -0.27 -2.21 0.97 12.70 -0.09 -0.93

6-7AM 0.47 5.06 0.15 2.74 -0.07 -1.03

8-9AM 0.89 10.16 0.05 0.96 0.05 0.75

10-11AM 1.02 11.71 0.17 3.43 0.13 2.20

12AM-1PM 1.12 12.69 0.26 5.22 0.14 2.26

2-3PM 1.02 11.44 0.11 2.09 0.21 3.46

4-5PM 1.06 11.89 0.06 1.18 0.22 3.60

6-7PM 0.93 10.35 0.25 5.00 0.15 2.37

8-9PM 0.54 5.87 -0.02 -0.32 -0.05 -0.83

10-11PM 0.38 4.02 0.12 2.25 -0.18 -2.64

Tuesday -0.06 -1.02 0.01 0.39 -0.07 -2.49

Wednesday -0.09 -1.54 0.18 6.97 -0.04 -1.42

Thursday -0.11 -1.89 0.05 1.82 -0.13 -4.18

Friday -0.15 -2.50 -0.15 -5.67 -0.01 -0.20

Saturday -0.47 -7.38 -0.18 -6.26 0.12 3.34

Sunday -0.67 -9.80 -0.02 -0.58 -0.02 -0.40

Holiday -0.73 -6.14 -0.14 -2.15 0.19 2.61

April, 2023 -0.26 -5.62 0.29 14.82 0.08 3.39

May, 2023 -0.19 -3.55 0.40 16.51 0.11 3.68

June, 2023 -0.32 -5.42 0.53 19.51 0.17 5.25

Figure 5: Feature importance for hourly trip count models

Examining the Impact of Weather, Temporal Factors, and User Traits on Multimodal Shared Micromobility Systems in Non-Urban Campus

Environments: The MORE Sharing Case Study

199

Analysis of the hourly trip rates throughout the

day (Table 3) revealed an increase in trip counts

between 6 AM and 11 AM compared to the period

between 0 and 2 AM. This indicates higher demand

during the morning hours, likely related to

commuting and the start of daily activities. Most

bookings occurred between 12 PM and 5 PM,

reflecting lunchtime and mid-afternoon activities.

After 6 PM, the trip rate started to decline, suggesting

reduced campus activity during the evening hours.

Significantly fewer trips were observed on Fridays,

weekends, and holidays than on Mondays. This pattern

suggests that MORE Sharing experiences reduced

usage during non-working days and weekends when

the campus community may have fewer obligations or

different transportation needs. The first month of

MORE Sharing operations also had more hourly

reservations than subsequent months.

When modeling hourly trip counts, the Random

Forest (RF), Gradient Boosted Regression Trees

(GBRTs), and Neural Networks (NN) models

exhibited similar performance but varied in feature

importance rates (Figure 5). RF placed considerable

importance on parameters such as the time of day

vehicle type, weekday type, and month when

predicting hourly trip counts. GBRTs also considered

vehicle type and time-related parameters as priorities.

NN assigned the highest importance to vehicle type,

followed by the wind chill index, day type, and

month. These disparities in feature importance

highlight the factors influencing trip counts and the

potential of different models to capture and predict

these patterns.

To estimate reservation durations and traveled

distances, additional parameters like pre-reservation

duration in minutes, user age, and the length of user

registration on MORE Sharing were included (Table

3). The NBR model indicated that these factors

played a minor role in predicting reservation duration

in minutes (Table 3). However, reservations made

during rainy weather tended to last longer, suggesting

that users prefer to reserve vehicles for extended

periods when weather conditions are unfavorable,

possibly to allow for flexibility or contingency plans.

E-cargo bikes had longer reservation durations

compared to e-bikes, city bikes, or e-scooters,

implying that users may require more time when

using e-cargo bikes, possibly due to the nature of

transporting goods or engaging in activities that

necessitate the use of e-cargo bikes. E-mopeds had

the longest reservations, potentially for more

extensive trips or specific purposes (Table 3).

Reservations between 3 and 5 AM consistently

exhibited the longest durations, and daytime

reservations generally surpassed those between

midnight and 2 AM (Table 3). Furthermore,

reservations on Fridays, weekends, and holidays

tended to be shorter than on Mondays. The duration

of reservations continuously increased since March

2023, with each successive month showing higher

average reservation durations.

While the Negative Binomial Regression (NBR)

model suggested only minor effects of user age, the

length of user registration, and pre-reservation

duration on predicting reservation durations (Table

3), both the Random Forest (RF) and Gradient

Boosted Regression Trees (GBRTs) effectively

utilized these parameters in predicting reservation

durations (Figure 6). Weather-related parameters

such as humidity and wind chill index also played a

role in predictions, with the wind chill index

demonstrating slightly higher feature importance than

wind speed and temperature parameters alone.

Vehicle type proved to be a relevant factor for RF and

GBRTs in predicting reservation duration, suggesting

that different vehicle types may exhibit distinct usage

patterns that influence the duration of reservations. In

contrast, the Neural Network (NN) model did not

reveal particular features important for predicting

reservation durations.

Concerning driven distances, users who pre-

reserved vehicles tended to cover longer distances

compared to those who made spontaneous reservations

(Table 3). This suggests that individuals who plan their

trips in advance may have specific destinations or

longer journeys in mind, resulting in increased

distances traveled. Additionally, female users traveled

longer distances than their male counterparts,

indicating potential variations in travel patterns,

purposes, or preferences within the MORE Sharing

system. Further exploration is necessary to elucidate

the underlying factors driving these differences.

Distances traveled were generally shorter during

rainy weather conditions (Table 3). This phenomenon

could be attributed to users opting for shorter trips or

seeking sheltered transportation alternatives during

adverse weather. E-cargo bikes were consistently

used for longer distances than city bikes,

underscoring their suitability for transporting goods

or engaging in activities necessitating extended

travel. The longest distances were covered by e-

mopeds, suggesting that users opt for e-mopeds when

they require swift and substantial distance coverage.

These results highlight the advantages of

incorporating multiple vehicle types within a shared

mobility system, as each type caters to distinct

purposes and accommodates a broad spectrum of

travel needs.

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

200

Distances traveled were higher during daytime

and evening hours than nighttime (Table 3), aligning

with typical travel patterns as users engage in various

daytime activities. On Tuesdays and Thursdays, trips

covered shorter distances, potentially reflecting

specific weekday routines or shorter commutes.

Conversely, trips on weekends and holidays tended to

be more extensive, implying that users engage in

extended leisure activities during these periods.

Overall, the length of trips in kilometers has exhibited

a steady increase since March 2023. Factors such as

increased familiarity with the service, expanded

usage scenarios, or evolving user preferences may

contribute to this observed growth.

RF and GBRTs underscored the significance of

vehicle type and the duration of user registration on

the MORE Sharing platform in predicting the

distance traveled (Figure 7). The wind chill index

emerged as a significant distance predictor,

representing the combined influence of temperature

and wind speed. Additionally, GBRTs factored in

whether the vehicle was pre-reserved, signifying the

impact of this factor on the distance traveled.

Figure 6: Feature importance for reservation duration models.

Figure 7: Feature importance for trip distance models

Examining the Impact of Weather, Temporal Factors, and User Traits on Multimodal Shared Micromobility Systems in Non-Urban Campus

Environments: The MORE Sharing Case Study

201

5 CONCLUSIONS

While shared micromobility systems in urban areas

have been extensively researched, these systems still

need to be studied more in non-urban settings and

smaller contexts, such as university or business

campuses, residential neighborhoods, and non-urban

communities. This study aims to bridge this

knowledge gap by investigating the multimodal

shared micromobility service at the University of the

Bundeswehr in Munich. This unique service allows

users to access multiple micromobility modes

through a single app and offers a complimentary

mobility budget.

Our research focused on tracking hourly trip

metrics, including trip counts, trip lengths, and

reservation times, during the initial four months of

this campus-based micromobility service. We

considered factors, including weather conditions,

time of day and week, lead times for reservations,

user demographics, and various vehicle types such as

shared city bikes, e-bikes, e-cargo bikes, e-scooters,

and e-mopeds. To gain insights into usage patterns,

we employed several machine learning models,

including Negative Binomial Regression (NBR),

Random Forests (RF), Gradient Boosted Regression

Trees (GBRTs), and Neural Networks (NN).

Our findings revealed diverse user engagement

levels, behaviors, and preferences within the MORE

Sharing service. E-scooters emerged as the most

frequently used vehicle type, closely followed by e-

bikes. E-cargo bikes and city bikes exhibited similar

usage rates, while e-mopeds had the lowest average

usage. However, it is essential to note that lower

average usage of e-mopeds does not equate to lower

popularity; users often reserve e-mopeds for extended

periods compared to other vehicle types. E-cargo

bikes follow a similar trend, with e-bikes and e-

scooters having comparable average reservation

times, while city bikes recorded the shortest

reservation durations. E-mopeds covered the most

extended distances, with e-bikes, e-cargo bikes, and

e-scooters showcasing similar average travel

distances, while city bikes covered the least distance.

These findings underscore the benefits of offering

multiple vehicle types within a shared mobility

system, catering to diverse travel needs and purposes.

Our research also highlighted that most trips,

regardless of vehicle type, occur on weekdays.

Furthermore, reservations on Fridays, weekends, and

holidays tend to be shorter in duration compared to

Mondays but involve longer distances during

weekends and holidays. The peak in the number of

trips typically falls between 6 AM and 8 AM, remains

high from 8 AM to 6 PM for all micromobility modes,

and decreases after 6 PM. Reservations between 10

AM and 7 PM generally exhibit the most extended

reservation times and travel distances.

We observed more hourly reservations during the

initial month of MORE Sharing operations than in

subsequent months. However, since March 2023,

both reservation durations and travel distances have

steadily increased, with each new month surpassing

the previous month's average reservation duration.

Rainy weather decreased the number of shared

micromobility trips and the distances traveled.

However, during rainy periods, reservations tend to

last longer. Users who reserve vehicles in advance

tend to cover greater distances than those who make

spontaneous reservations. On average, female users

travel farther than their male counterparts.

This research underscores the significance of

understanding local contexts and community needs

when implementing shared micromobility systems in

non-urban settings. Policymakers, urban planners,

and transportation providers can leverage these

insights to enhance the design and implementation of

shared micromobility systems in various

microenvironments, such as campuses, residential

neighborhoods, and corporate settings. This study

contributes to the broader objective of promoting

sustainable transportation initiatives and

environmentally friendly mobility options across

diverse settings. Examining and optimizing shared

micromobility solutions in various contexts can pave

the way for more efficient and accessible mobility

solutions.

ACKNOWLEDGEMENTS

This research is part of the project MORE – Munich

Mobility Research Campus. The project is funded by

dtec.bw – Digitalization and Technology Research

Center of the Bundeswehr. dtec.bw is funded by the

European Union –NextGenerationEU. We want to

thank eVhcle (Richard Kemmerzehl, Christoph

Ulusoy, Bastian Biener, Constantin Mossgraber, and

Zeeshan Zahoor) for their successful cooperation in

the MORE Sharing project.

REFERENCES

Analytics Vidhya. (2023). Regression Analysis Using

Artificial Neural Networks.

Becker, S., & Rudolf, C. (2018). Exploring the Potential of

Free Cargo-Bikesharing for Sustainable Mobility.

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

202

GAIA - Ecological Perspectives for Science and

Society, 27(1), Article 1. https://doi.org/10.14512/

gaia.27.1.11

Breiman, L. (2001). Random Forests (45). 45, Article 45.

https://doi.org/10.1023/A:1010933404324

Davis, J. (2014). Gradient Boosted Regression Trees for

Forecasting Daily Solar Irradiance from a Numerical

Weather Prediction Grid Interpolated with Ordinary

Kriging.

DWD. (2023). Weather and Climate in Munich.

evhcle. (2023). FAQs MORE Sharing. https://evhcle.com/

faqmoreshare

Fishman, E., Washington, S., & Haworth, N. (2015).

Bikeshare’s impact on active travel: Evidence from the

United States, Great Britain, and Australia. Journal of

Transport & Health, 2. https://doi.org/10.1016/

j.jth.2015.03.004

Fishman, E., Washington, S., Haworth, N., & Watson, A.

(2015). Factors influencing bike share membership: An

analysis of Melbourne and Brisbane. Transportation

Research Part A Policy and Practice, 71, 17–30.

https://doi.org/10.1016/j.tra.2014.10.021

Gebhart, K., & Noland, R. (2014). The impact of weather

conditions on bikeshare trips in Washington, DC.

41(6), 1205–1225. https://ideas.repec.org/a/kap/transp/

v41y2014i6p1205-1225.html

Groen, G. (2009). Wind chill equivalent temperature

(WCET) Climatology and scenarios for Schiphol

Airport.

Huber, R. (2023). Mobilität der Zukunft: Der richtige Mix

ist entscheinend. Mobile Faszination, 3.

MORE Sharing. (2023). MORE Sharing für die Universität

der Bundeswehr.

NACTO. (2022). Shared Micromobility in the US: 2020-

2021. https://nacto.org/wp-content/uploads/2022/12/

2020-2021_shared_micro_snapshot_Dec7_2022.pdf

Noland, R. B. (2021). Scootin’ in the Rain: Does Weather

Affect Micromobility? Transportation Research Part

A: Policy and Practice, 149, 114–123. https://doi.org/

10.1016/j.tra.2021.05.003

Pobudzei, M., Kemmerzehl, R., Ulusoy, C., & Hoffmann,

S. (2023). Beyond the City: Testing Shared

Micromobility with MORE Sharing. 25th Euro Working

Group on Transportation Meeting (EWGT 2023),

Santander.

Pobudzei, M., Sellaouti, A., Tießler, M., & Hoffmann, S.

(2022). Riders on the Storm: Exploring Meteorological

and Temporal Impacts on Shared E-Scooters (SES) in

Munich, Germany. 1–6. https://doi.org/10.1109/

ISC255366.2022.9922429

Pobudzei, M., Wichmann, I., & Hoffmann, S. (2023).

Unlocking the Wheel: Insights into Shared

Micromobility Perceptions and Adoption on Campus.

9th International Conference on Vehicle Technology

and Intelligent Transport Systems (VEHITS 2023).

Rérat, P. (2021). The rise of the e-bike: Towards an

extension of the practice of cycling? Mobilities, 16(3),

Article 3. https://doi.org/10.1080/17450101.2021.1897

236

Schreier, H., Sellaouti, A., Tiessler, M., Pobudzei, M.,

Hoffmann, S., Hager, A., Hericks, K., Grimm, C., &

Brümmer, M. (2022). Evaluierung der verkehrlichen

Wirkungen von E-Tretrollern. Landeshauptstadt

München.

https://muenchenunterwegs.de/content/1423/download

/220530-bericht-eva-et-final-web.pdf

Tießler, M., Pobudzei, M., Sellaouti, A., & Hoffmann, S.

(2023). Electric Scooters and Where to Find Them—A

Spatial Analysis of the Utilization of Shared E-Scooter

in Munich, Germany.

UniBw. (2023a). Die zivile Gleichstellungsstelle der UniBw

M. Gleichstellung. https://www.unibw.de/gleich

stellung

UniBw. (2023b). Karriere an der Universität der

Bundeswehr München.

Younes, H., Zou, Z., Wu, J., & Baiocchi, G. (2020).

Comparing the Temporal Determinants of Dockless

Scooter-share and Station-based Bike-share in

Washington, D.C. Transportation Research Part A:

Policy and Practice, 134, 308–320. https://doi.org/

10.1016/j.tra.2020.02.021

Examining the Impact of Weather, Temporal Factors, and User Traits on Multimodal Shared Micromobility Systems in Non-Urban Campus

Environments: The MORE Sharing Case Study

203