The project was conducted as part of a graduate statistical consulting course at Portland State University in collaboration with TREC (Transportation Research and Education Center) across a 10-week Winter 2025 term (January - March 2025). The goal was to compare two different datasets, but the scope shifted to focus more on the data integration and cleaning process.

Executive Summary

This project explored two primary sources of traffic data—PORTAL (managed by ODOT and WSDOT) and INRIX (a commercial provider)—to investigate travel time trends on highways in the Portland metropolitan area. Each dataset presents unique strengths and limitations, ranging from sensor coverage and data frequency to the way road segments are defined. Integrating the datasets required a robust approach to handling differences in geometry (e.g., segment boundaries), time intervals, and data quality or availability.

A core portion of the work involved "cleaning" and "harmonizing" the two datasets so that they could be combined in a single framework. This included transforming hex-encoded road segments (in PORTAL) into usable geometry objects, standardizing coordinate reference systems, normalizing timestamps across multiple years, and filtering records to ensure consistency in coverage. Ultimately, the merged dataset enables more comprehensive analysis—such as year-to-year trend detection but also potential pitfalls in the data, such as not enough coverage between the segments in PORTAL and INRIX.

Key achievements include

• Metadata Integration: Successfully joined PORTAL station segments with INRIX TMC (Traffic Message Channel) segments via spatial intersection. This step highlighted how different segmentation tation and directional labeling can lead to data mismatches.
• Time-Series Standardization: Unified 15-minute interval data from both sources, converting timestamps to a consistent time zone and format for seamless comparison.
• Validation and Visualization: Used GeoPandas plotting to confirm that station coverage areas correctly overlapped with TMC segments. This allowed identification of potential misalignment or missing station geometry.
• Foundations for Incident Analysis: While the scope shifted toward data-integration workflows, the cleaned and merged dataset can now be readily used to overlay incident data and evaluate how quickly each source detects and captures major congestion.

Code and Report

Code located on GitHub: whitham-powell/cades-traveltime-compare

Link to the final report: CADES Travel Time Comparison Report (PDF)