Seattle Micromobility: Spatiotemporal Analysis, Safety Hotspots & Geographic Influences 🚲🛴

Group members: Abdul-Razak Alidu, Sruangsaeng Chaikasetsin, Hunter Lybbert, Ysabel Yu

---

Table of Contents

1. Summary
2. Introductory Background
3. Problem Statement & Objectives
4. Datasets
5. Tools & Packages
6. Methodology & Approach
7. Results, Conclusions, and Future Directions
8. References

---

1. Summary

This project investigates micromobility usage-specifically e-scooters and bike share in Seattle. Our primary focus is to understand where and when people ride, locate potential collision hotspots involving vulnerable road users (pedestrians, cyclists, e-scooter riders), and examine geographic factors (like slope, transit proximity, land use) that may influence both usage and collisions. We hope these insights will guide data-driven enhancements to infrastructure, user behavior, and environmental considerations- ultimately fostering safer and more sustainable micromobility adoption.

Contents of the repo:

• notebooks contains our jupyter notebooks where a majority of the data analysis we performed took place. There are subdirectories for each aspect of the work: bikeshare analysis, collision analysis, and raster based analysis
• visuals contains the many visualizations that we have created for this project.
• sample_data contains a collection of samples of the data we used. Not all of the data could be included due to size constraints.
• scripts contains the scripts used in gathering and processing the bikeshare data.

---

2. Introductory Background

Seattle has rapidly expanded its micromobility offerings (e-scooters, bike shares) to provide low-emission, last-mile travel solutions. For example, Seattle's e-scooter pilot program recorded over 1.4 millions rides between 2020 and 2021 (Seattle Department of Transportation (SDOT), 2021). Meanwhile, a 2022 study by the National Association of City Transportation Officials (NACTO) found that micromobility devices can replace up to 30% of short car trips in dense urban areas, emphasizing their role in reducing vehicle miles traveled (NACTO, 2022). However, questions about safety and usage distribution persist. According to SDOT collision data, 21% of reported e-scooter crashes in 2021 led to serious injuries-indicating the need for improved infrastructure and rider education (SDOT, 2022). Additionally, steep slopes, uneven access to protected lanes, and transit deserts may further complicate safe micromobility usage for residents in certain neighborhoods (King County Mobility Coalition, 2020) By merging trip data (where and when people ride) with collision records and relevant geospatial layers (slope, transit stops, land-use types), we aim to

• Pinpoint high-usage corridors and collision hotspots
• Explore geographic influences (e.g., terrain, neighborhood zoning) on ridership and collisions

---

3. Problem Statement & Objectives

Problem Statement

1. Where do e-scooter and bike share trips cluster, and do these align with collision-prone zones?
2. How do factors like slope, proximity to infracstructure, or land use affect micromobility usage and crash risk?

Objectives

1. Spatiotemporal Patterns
   • Identify ridership peaks (daily/weekly/seasonal) and produce usage hotspot maps.
2. Collision Hotspots
   • Detect VRU collision clusters using Kernel Density Estimation (KDE) to compare with usage patterns.
3. Geographic Factors
   • Integrate slope or land-use data to see how terrain/environment shapes micromobility usage and collisions.

---

4. Datasets

1. Micromobility Trip Data 🚲🛴
   General Bikeshare Feed Specification (GBFS) – Standardized format for bike/scooter share data
   🔗 GitHub Repository

   • Bird Feed
   • Lime Feed

2. Seattle Micromobility & Collision Data 🏙️
   Seattle Open GIS Portal
   🔗 Seattle GIS Open Data

   • 🚦 Collision Data
     • Seattle SDOT Collisions (All Years)
   • 🚲 Infrastructure & Facilities
     • SDOT Bike Facilities
     • Bicycle Racks
     • Sidewalks
   • Seattle Open Data Portal

3. Geographic & Slope Data 🌍

   • 🔗 Washington State DNR LiDAR Portal – 2016 King Country high-resolution LiDAR elevation data for slope analysis

---

5. Tools & Packages

• Tools/packages used
  • Geopandas
  • Numpy
  • Matplotlib
  • Rasterio
  • Rioxarray
  • Xarray
  • Rasterstats
  • Contextily
  • bikeshare-client-python to access (GBFS)

---

6. Methodology & Approach

Spatiotemporal Patterns: Bikeshare counter data was plotted on maps by Seattle neighborhood and turned into animations to show bikeshare changes over time. Scooter and bike distribbutions were plotted using Kernel Density Estimation to find micromobility use hotspots. This data was also used to do a time series analysis per hour and per day to find patterns in usage.

Collision Hotspots: For our collision hotspot analysis, we examined Seattle Department of Transportation (SDOT) collision data from 2020 to February 2025, focusing on crashes involving vulnerable road users (VRUs) such as pedestrians, cyclists, and scooter riders. We began by cleaning the dataset—removing duplicates, addressing missing values, converting date-time fields for consistency, and setting the appropriate coordinate reference system (CRS) for geospatial work. Next, we conducted exploratory analysis to tally collisions by year and assess severity, finding that 21% resulted in serious injuries. To uncover spatiotemporal patterns, we mapped crashes by hour, day, and month, then applied Kernel Density Estimation (KDE) to identify hotspot locations. We also performed a proximity analysis using GeoPandas’ sjoin_nearest function to determine how many collisions occurred within 50 meters of bike facilities and sidewalks, revealing infrastructure-related risks. Finally, we calculated per-capita collision rates by neighborhood, normalizing crash counts per 1000 residents to highlight areas with disproportionate risk. This multi-step approach, detailed in notebooks/collision, anchors our safety insights.

Geospatial Correlation: Geospatial correlation methods were split into 2 main parts: raster manipulation and road zonal stats. Firstly, a full Seattle DTM raster was created from smaller raster sets provided by King County. From this a slope raster was created and used to calculate zonal stats for roads in U District and Downtown. The stats showed the mean slope of those roads and highlighted where slopes were greatest using plots.

7. Results, Conclusions, and Future Directions

Conclusion

Our analysis revealed key safety insights for Seattle’s micromobility landscape.

• Micromobility Usage: High-usage zones like Downtown and U District often correlate with more crashes
• Safety Gaps: High-usage zones like Downtown often correlate with more crashes, but infrastructure plays a critical role—bike facilities show higher risks (~60% of collisions) compared to multi-used trails (~5%), likely due to conflict with the vehicle traffic.
• Geographic Factor: Slope is not an indicator for any of the above.
• Limits: A key limitation is that collision data lacks real-time trip counts, making usage overlap estimates approximate rather than precise.

Future Directions

To build on this work, we propose several next steps:

• Normalize collision rates with actual trip data (not just availability) for a clearer usage-crash link.
• Expand land use analysis to include factors like transit stops and commercial zones, which may further influence risk.
• Refine hotspot detection using DBSCAN to identify tighter, more actionable clusters compared to KDE’s broader smoothing.

8. References

1. Seattle Department of Transportation (SDOT).** (2021). *E-Scooter Pilot Program Evaluation Report. Seattle, WA.
2. National Association of City Transportation Officials (NACTO). (2022). Shared Micromobility in the U.S.: 2022 Snapshot. https://nacto.org
3. Seattle Department of Transportation (SDOT). (2022). Collision Data Dashboard. https://www.seattle.gov/transportation/
4. King County Mobility Coalition. (2020). Mobility Needs Assessment for King County. King County, WA.