Agenda
Notes
Overview
Joe Flood presented the planned approach for developing the Application and Analysis Guide as part of Phase 11C. The guide will cover how to set up, run, and analyze results for three ActivitySim application scenarios, using the SANDAG example model as the reference implementation. The guide will be built in Markdown and is intended as a living starting point, extensible in future phases.
Application Scenarios
Each scenario will be structured in three parts: (1) preparing the scenario, (2) running it, and (3) analyzing the results. Code blocks and links to Python notebooks will be included throughout.
1. Land Use Change
A densification scenario around a transit-oriented neighborhood will be used to illustrate how to modify land use files (population, employment fields). An alternate synthetic population will be needed; Joe may request a PopulationSim run from SANDAG modelers. Key metrics include tour mode share and average tour length by purpose.
2. Network Change
Focused on the addition of a new transit line. Changes on the ActivitySim side are limited to updating skim files; guidance will be provided on where to download and place updated skims. Joe may request updated transit skims from SANDAG. Key metrics include tour mode share, travel times, and jobs accessible within a 30–60 minute transit ride.
3. Telecommuting Change
Requested by the consortium in response to post-COVID changes in work patterns. Joe Flood planned to begin this scenario after WSP completes their explicit telecommuting model work. However, Joe Castiglione noted that Phase 11 only funds 2 of WSP's 4 identified model enhancements, so the full telecommute suite will not be available. Consensus was to use existing model capabilities rather than waiting on new development. Key metrics include CDAP shares, VMT, and transit trips.
Key Discussion Points
Documentation Structure
There was significant discussion about whether the guide should be a third top-level document (alongside the existing User Guide and Developer Guide), or nested within the User Guide. Marty Milkovits advocated for a separate guide targeting senior modelers and managers — focusing on when and why to apply the model, not just the mechanics. Sumit argued that practitioners would benefit from having everything in one place under the User Guide. Joe Flood leaned toward a third standalone guide but acknowledged discoverability is the most important factor. No final decision was made; the topic will continue asynchronously.
AI and Documentation
Amir (TfNSW) raised the potential for AI-assisted documentation, sharing that TfNSW pushed Sphinx documentation to Confluence and used its built-in AI chatbot with low risk of hallucinations. Joel Freedman noted that AI tools are already capable of answering mechanical "how to" questions, but struggle with contextual guidance (e.g., model suitability, limitations, and appropriate use cases). The group agreed AI integration is out of scope for now but worth exploring in future phases.
Scenario QA
A question from the chat raised whether the guide would include quality assurance steps to confirm scenarios ran as intended. Joe Flood agreed this was valuable and noted he could add a brief pre-analysis check to verify input changes were applied correctly before reviewing outputs.
Action Items
| Owner |
Action |
Deadline |
| Joe Flood |
Create GitHub Discussion threads (one for overall approach + one per scenario) for async feedback |
Today (April 7) |
| Consortium members |
Provide initial feedback on GitHub Discussions |
April 21) |
| Joe Castiglione |
Agendize GitHub feedback check-in at the April 21 meeting |
April 21 |
| Joe Flood |
Coordinate with SANDAG modelers regarding scenario inputs |
Ongoing |
| Joe Flood |
Coordinate with WSP on telecommuting model status as necessary |
Ongoing |
| Shaun Tabone |
Present summary of consortium agency visualization tools survey |
April 21 |
| Joe Castiglione |
Create visualization subdirectory in shared drive for Shaun's materials |
Soon |
Agenda
Notes
Overview
Joe Flood presented the planned approach for developing the Application and Analysis Guide as part of Phase 11C. The guide will cover how to set up, run, and analyze results for three ActivitySim application scenarios, using the SANDAG example model as the reference implementation. The guide will be built in Markdown and is intended as a living starting point, extensible in future phases.
Application Scenarios
Each scenario will be structured in three parts: (1) preparing the scenario, (2) running it, and (3) analyzing the results. Code blocks and links to Python notebooks will be included throughout.
1. Land Use Change
A densification scenario around a transit-oriented neighborhood will be used to illustrate how to modify land use files (population, employment fields). An alternate synthetic population will be needed; Joe may request a PopulationSim run from SANDAG modelers. Key metrics include tour mode share and average tour length by purpose.
2. Network Change
Focused on the addition of a new transit line. Changes on the ActivitySim side are limited to updating skim files; guidance will be provided on where to download and place updated skims. Joe may request updated transit skims from SANDAG. Key metrics include tour mode share, travel times, and jobs accessible within a 30–60 minute transit ride.
3. Telecommuting Change
Requested by the consortium in response to post-COVID changes in work patterns. Joe Flood planned to begin this scenario after WSP completes their explicit telecommuting model work. However, Joe Castiglione noted that Phase 11 only funds 2 of WSP's 4 identified model enhancements, so the full telecommute suite will not be available. Consensus was to use existing model capabilities rather than waiting on new development. Key metrics include CDAP shares, VMT, and transit trips.
Key Discussion Points
Documentation Structure
There was significant discussion about whether the guide should be a third top-level document (alongside the existing User Guide and Developer Guide), or nested within the User Guide. Marty Milkovits advocated for a separate guide targeting senior modelers and managers — focusing on when and why to apply the model, not just the mechanics. Sumit argued that practitioners would benefit from having everything in one place under the User Guide. Joe Flood leaned toward a third standalone guide but acknowledged discoverability is the most important factor. No final decision was made; the topic will continue asynchronously.
AI and Documentation
Amir (TfNSW) raised the potential for AI-assisted documentation, sharing that TfNSW pushed Sphinx documentation to Confluence and used its built-in AI chatbot with low risk of hallucinations. Joel Freedman noted that AI tools are already capable of answering mechanical "how to" questions, but struggle with contextual guidance (e.g., model suitability, limitations, and appropriate use cases). The group agreed AI integration is out of scope for now but worth exploring in future phases.
Scenario QA
A question from the chat raised whether the guide would include quality assurance steps to confirm scenarios ran as intended. Joe Flood agreed this was valuable and noted he could add a brief pre-analysis check to verify input changes were applied correctly before reviewing outputs.
Action Items