What factors contribute to e-scooter crashes: A first look using a naturalistic riding approach.

INTRODUCTION: Shared dockless electric scooters (e-scooters) are a popular shared mobility service providing an accessible last-mile transportation option in urban and campus environments. However, city and campus stakeholders may hesitate to introduce these scooters due to safety concerns. While prior e-scooter safety studies have collected injury data from hospitals or riding data under controlled or naturalistic conditions, these datasets are limited and did not identify risk factors associated with e-scooter riding safety. To address this gap in e-scooter safety research, this study collected the largest naturalistic e-scooter dataset to date and quantified the safety risks associated with behavioral, infrastructure, and environmental factors. METHOD: A fleet of 200 e-scooters was deployed on Virginia Tech's campus in Blacksburg, VA for a 6-month period. Fifty were equipped with a unique onboard data acquisition system, using sensors and video to capture e-scooter trips in their entirety. The resulting dataset consisted of 3,500 hours of data spanning over 8,500 trips. Algorithms were developed to identify safety critical events (SCEs) in the dataset and analyses were conducted to determine the prevalence of various SCE risk factors and associated odds ratios. RESULTS: Results from this study indicate that infrastructure-related factors, behavior of e-scooter riders and other actors, and environmental factors all contributed to the SCE risk for e-scooter riders in Virginia Tech's pedestrian-dense campus environment. CONCLUSIONS: To help mitigate unsafe rider behavior, educational outreach programs should quantify the significant risks associated with infrastructure, behavioral, and environmental risk factors and provide clear recommendations to riders. Improved infrastructure maintenance and design may also improve safety for e-scooter riders. PRACTICAL APPLICATIONS: The infrastructure, behavioral, and environmental risk factors quantified in this study can be applied by e-scooter service providers, municipalities, and campus administrators to develop mitigation strategies to reduce the safety risks associated with e-scooter deployments in the future.

Estimating small area demand for online package delivery.

Using publicly available microdata sets, we show how estimates for online delivery purchases can be generated for small geographic areas defined in our study as micro analysis zones (MAZ) and how these estimates vary across the MAZs that featured in our study. With a focus on Miami-Dade County, we use both the national household travel survey (NHTS) data and synthetic data obtained from Southeast Florida Regional Planning Model (SERPM) to generate demand estimates of online delivery purchases for more than 5300 distinct geographic units in Miami-Dade County. We assess the quality of the estimates using measures of predictive accuracy and by comparing the cumulative values obtained with the population estimates generated from the NHTS survey data for Miami-Dade County. Our approach fills a void in the area of purchases of online delivery items where rich observable data are typically unavailable and it also provides the added potential benefit of being easily replicated nationwide given the emphasis on the use of publicly available data.