Land Use Change Guide Wishlist

[JoeJimFlood]

This discussion is for expressing desires for the Land Use Change scenario, including specifics on the scenario to test and what metrics to demonstrate the calculation of. It is proposed to increase the density of a neighborhood as that is a common contemporary travel model application. Suggestions from the April 7 consortium included demonstrating how to split zones and reallocating existing population as opposed to adding population. Some proposed metrics to demonstrate the calculate of include:

• Tour mode share
• Average tour length by purpose

Please submit any comments before the April 21 consortium meeting.

[sumitbindra]

Land-Use Scenario

Problem Statement

1. (TransLink preference) A new TOD development with an additional 2,000 residential units and 1,000 retail jobs (mixed land-use scenario) near a light rail station.

   • Do auto ownership rates for new households drop compared to similar households elsewhere in the region?
   • Does the model proportionally reflect the increased attraction for those zones?

2. What does student enrollment look like if a new school is opened? How does it impact escorting trips?

Components Impacted (Edit based on modeling system for testing)

1. Synthetic Population and Land Use Input Files – Land use changes on households and/or population requires regeneration of synthetic population to reflect the changes in households in the TOD development areas.

2. Destination Choice Size Terms – Land use changes directly affect the size terms in destination choice. The guide should walk through how employment and household counts flow into those size term expressions, and how a modeler can sanity-check that the model is actually "seeing" the land use change in destination choice behavior.

3. Mode Shift – In the TOD scenario, we should observe a shift towards active and transit mode and the guide should demonstrate how to summarize that.

4. Accessibility Logsums – These feed upstream into auto ownership and other submodels. A TOD scenario is a great case because improved accessibility near the station should be suppressing auto ownership for nearby households.

Outputs / Summary

1. VMT
2. Mode shift
3. Active mode share near the development
4. Auto ownership

[amsamimi]

A TOD scenario is a strong candidate, with a few considerations. For alternative land use, PopulationSim’s repop mode should be used to avoid synthesis noise affecting results elsewhere. In a typical TOD scenario, total population should remain constant, with growth in target zones offset by redistribution from other TAZs (e.g., via a global reduction factor). Maintaining total population is important for comparability. While not currently supported in ActivitySim, it would be useful to enable or document a way to enforce population totals at the TAZ level, since PopulationSim controls households but not population.

[JoeJimFlood]

I did see that comment. My concern isn't regarding adding to the total population vs reallocating households from elsewhere. It also says in the scope "A step-by-step tutorial using the SANDAG ABM3 Example that users can follow along with will also be included." Such a tutorial is important to me as I feel like it is the sort of thing I personally would find very useful. My concern is that adding PopulationSim in repop mode will add a lot of effort in setting up and running that tutorial (which I would like to test on my end). This is why I was suggesting writing a section on using PopulationSim but then for the tutorial just providing a quick script for users to edit the input files.

[amsamimi]

Hi @JoeJimFlood
If you haven’t used repop mode before, I’d be happy to walk you through it. For this task, I don’t think we need to create a full new user guide for PopulationSim repop; we can simply frame it as a step within testing altered land-use scenarios. Enhancing the PopulationSim guide could be a future task, but the existing documentation should be sufficient for our current purpose.

My main concern with writing a custom Python script to relocate households across selected travel zones is that we would lose control over the demographic profile. That approach is not intentionally designed and contradicts most standard alternative land-use scenario testing practices. For this task, we shouldn’t endorse a method that is theoretically weak, especially for such an important use case.

[JoeJimFlood]

@amsamimi I appreciate the offer, but I've run the example and can figure it out. If we create a synthetic population using PopulationSim's prepop mode we can just have users download it, but it might be too large of a file if the tutorial uses the full data for the SANDAG example, which is why I was suggesting using a script for the demo (I could add a disclaimer that such an action would not be recommended in practice).

There is a cropped dataset (initially used for testing) available in the SANDAG ABM3 Example that includes just Downtown San Diego and some of the neighborhoods to the north of there. If the tutorial used that then alternate synthetic population files could easily be uploaded to GitHub with an explanation of how they were created (the runtimes would also be much lower). I did map that dataset and there is a light rail station (Middletown on the Green and Blue Lines) that would be a good candidate to demonstrate how to run a TOD scenario in ActivitySim (though not a good real-life candidate given its proximity to a major airport runway).

[amsamimi]

Thanks, @JoeJimFlood, this is helpful. Just to confirm, is your main concern the file size? If so, which files specifically?
I’m assuming you mean the synthetic population (households and persons). We’ve moved from CSV to Parquet, which greatly reduces size, for example, in the Sydney case the HH file is under 10 MB and the person file under 30 MB, which may make sharing pre-generated populations more practical for a tutorial.

[JoeJimFlood]

@amsamimi My primary concern is actually the additional time that setting up PopulationSim for this would require. I did just get access to SANDAG's data and configs and am thinking that it could take less time than I initially thought.

[JoeJimFlood]

I'm setting up a run that adds 2000 households and 1000 retail jobs in the vicinity of the Grossmont Trolley Station on the Orange and Green Lines, effectively transforming the nearby mall into a TOD (they actually just broke ground on redeveloping that mall, but the plans don't include housing). I'm writing briefly on how to adjust PopulationSim config files, but will be providing a script that moves households from outside of the study area into the study area, which both keeps the total population constant and means that additional input files won't need to be provided.

[amsamimi]

Is there any reason for not using PopulationSim’s repop mode?

[joecastiglione]

@JoeJimFlood, before proceeding to setting up a run, and given the limited budget for this task, I think it may be useful to first get a bit more input from agency staff to define and refine the details of any tests, and make sure there's consensus. I think the most efficient way to do this may be through starting the outline of the ActivitySim Application & Analysis Guide, which is the first deliverable. The outline of the land use change could be the start of this doc.

[sumitbindra]

@JoeJimFlood - Based on a discussion here is the update guidance for this task:

• The scenario should represent a TOD (Transit-Oriented Development) land use change—specifically, densification near a transit station.
• The scenario should build upon an exising SANDAG "baseline" model run, likely a post-COVID (2023?) baseline
• An increment of 2000 new households will be added to nearby (more than one) TAZs/MAZs in a TOD area layered/added on top of a baseline (not a redistribution from elsewhere in the region)
• The new household and population increment should be defined with a distinct demographic profile appropriate for TOD residents (e.g., younger households, couples without children, or young professionals—not extended families). In PopSim this involves identifying demographic donor zone(s).
• PopSim’s repop feature should be used to append (not replace) households to the target TAZ(s)
• An increment of 1,000 jobs should be added to nearby (more than one) TAZs/MAZs in a TOD area. The distribution of employment by industry sector should reflect the mix of employment in other successful TOD areas (ie. probably less than 100% retail) / donor zones.
• Because the population is being added (not redistributed), total regional population will differ between the baseline and the scenario. When reporting results, consider metrics that correct for this (eg. per-capita metrics, mode shares to work from TOD area residents vs non-TOD area residents)

Please let us know if you have additional questions.