Developing a Data Fusion Tool for Improved Traffic Crash Exposure Analysis and Modeling

Accurate measurement of exposure is critical for understanding and preventing traffic crashes, as crash frequency is directly related to how much road users are exposed to risk. However, current exposure estimates rely on data sources with complementary but individually insufficient characteristics. Traditional traffic counts and Annual Average Daily Traffic (AADT) offer high accuracy but limited spatial and temporal coverage, while emerging Location-Based Services (LBS) data provide high-resolution mobility patterns but are often biased and less reliable. This fundamental mismatch between accuracy and coverage prevents agencies from developing the complete and reliable exposure estimates needed for effective safety analysis and planning. This project develops a data fusion tool that integrates traffic counts and AADT, LBS data, and socio-demographically representative survey data from the National Household Travel Survey (NHTS) into a unified measure of exposure. Unlike previous efforts that focused on a single travel mode or low temporal resolution, the proposed framework generates exposure estimates for motor vehicles, pedestrians, bicyclists, and scooters at fine spatial scales (intersection and mid-block) and temporal scales (daily and monthly). The tool is evaluated in Washington, D.C., using three alternative fusion paradigms: Bayesian fusion through hierarchical or state-space modeling, Dempster–Shafer theory for explicit uncertainty representation and accommodation of LBS coverage gaps, and model-based fusion employing structured error modeling with NHTS socio-demographics to correct LBS data bias. The fusion methods are compared through crash prediction models estimated with fused exposure measures against models using individual data sources, evaluated via pseudo-R², AIC, BIC, and out-of-sample prediction error, with a target improvement of at least 10% in predictive performance. Fused exposure patterns are further validated against Washington, D.C.’s High Injury Network and independent ground-truth count data where available. The final tool is delivered as an open-source Python package with documentation and secure coding practices. Agency outreach, including engagement with D.C. stakeholders managing the High Injury Network, informs tool refinement and supports preparation for future pilot deployment. This research supports USDOT’s Safety priority by generating more accurate and complete multimodal exposure measures that enable better identification of high-risk locations, improved crash prediction, and targeted safety interventions

Language

• English

Project

• Status: Active
• Funding: $91,000.00
• Contract Numbers:

  69A3552348323

• Sponsor Organizations:

  Office of the Assistant Secretary for Research and Technology

  University Transportation Centers Program
  Department of Transportation
  Washington, DC  United States  20590
• Managing Organizations:

  Howard University

  2400 6th Street, NW
  Washington, DC  United States  20059
• Project Managers:

  Bruner, Britain

• Performing Organizations:

  Howard University

  2400 6th Street, NW
  Washington, DC  United States  20059
• Principal Investigators:

  Marin, Claudia

• Start Date: 20260202
• Expected Completion Date: 20261231
• Actual Completion Date: 0
• USDOT Program: University Transportation Centers Program

Subject/Index Terms

• TRT Terms: Annual average daily traffic; Crash risk forecasting; Data fusion; Location data; Traffic counts; Traffic crashes; Travel surveys
• Geographic Terms: Washington (District of Columbia)
• Subject Areas: Data and Information Technology; Highways; Pedestrians and Bicyclists; Planning and Forecasting; Safety and Human Factors;

Filing Info

• Accession Number: 01978544
• Record Type: Research project
• Source Agency: Research and Education for Promoting Safety (REPS) University Transportation Center
• Contract Numbers: 69A3552348323
• Files: UTC, RIP
• Created Date: Feb 3 2026 3:31PM