Abstract Manager

Across agencies, the bottleneck in transportation modelling is increasingly not the model itself, but the delivery of reliable results at the pace decisions demand. York Region operates an Activity-Based Model (ABM) to support long-range transportation planning, policy development, and infrastructure investment decisions. The ABM framework is computationally intensive, integrating multiple component models. York Region staff regularly need to validate and transfer gigabytes of data between component models. Maintaining data quality and delivering model results in a reasonable amount of time when dealing with such large volumes of data has become increasingly challenging.

To streamline data exchange between component models, York Region has developed a portfolio of Jupyter notebooks to automate data processing and validation. Originally developed during the latest Regional Transportation Master Plan update, these tools have been built up over time by different contributors and have evolved to be used as the Transportation Planning team’s main resource for delivering reliable and timely model results.

As the group’s volume of work has increased, the existing workflow has become cumbersome due to inconsistent coding practices, fragmented documentation, and manual intervention needed for error monitoring and validation. The Region also faced challenges in maintaining the code base due to limited staff availability and insufficient knowledge transfer between team members.  The Region engaged Arcadis’ Digital Asset Management consultants to review and upgrade the portfolio of Python code to offer a more cohesive, user-friendly, and reliable experience. Arcadis documented existing workflows performed by Region staff, identified pain points, and developed a style guide to standardize Jupyter Notebook structure, Python code style, Markdown usage, visualization practices, data management, and reproducibility protocols. York Region and Arcadis are currently enhancing the Python codebase to adhere to the style guide. 

This presentation will offer the following insights, aimed at supporting peer agencies’ ability to deliver reliable model results in time and resource constrained situations:

• Lessons learned from operating a sophisticated travel demand model in an organization with numerous internal and external clients and a diverse portfolio of use cases;
• Techniques for automating the execution and validation of model results; and
• Guidelines for managing Python codebases used for transportation data processing and analysis.

Attendees will leave with practical, transferable methods to reduce friction, shorten turnaround, and improve reliability, so modeling teams can spend more time on delivering insights and less on data management, regardless of the specific platform they use.