About
A civic-research project that reads every Toronto park through two complementary lenses, Jane Jacobs’ urban vitality and an ecological-comfort framework, and makes the tradeoffs between them visible.
Origin: re-reading Jane Jacobs
This project started with a re-read. In early 2026, after fifteen years of carrying the book around mostly as a reference, I sat back down with Jane Jacobs’ The Death and Life of Great American Cities. The first reading, fifteen years ago, had stuck with me as a set of slogans: “eyes on the street,” “mixed primary uses,” “short blocks.” The second reading was different. The argument is much sharper than the slogans suggest, and Jacobs’ chapter on parks specifically (chapter 5) is more skeptical and more demanding than I remembered. She does not think parks are inherently good. She thinks neighbourhood parks are created by the urban form around them. The same patch of green can be a treasured living room or a feared deadzone depending entirely on what presses up against it.
That sat badly with the way Toronto talks about its parks. We rank them, we celebrate the iconic ones, we lament that there aren’t more of them, and we mostly ignore the question Jacobs was asking: is this park doing the urban work parks are supposed to do? The atlas is one attempt to make that question askable at the scale of the whole city.
What this project is
An interactive atlas of Toronto’s 3,273 mapped parks. Every park is scored across six dimensions (Edge Activation, Connectivity, Amenity Diversity, Natural Comfort, Enclosure, Border Vacuum Risk), classified into one of twelve typologies, clustered into eight groups, and described with an auto-generated narrative that names its strengths, weaknesses, and tradeoffs.
The data comes from City of Toronto Open Data (Green Spaces, Parks & Recreation Facilities, Centreline V2, Pedestrian Network, Treed Area, Ravine Bylaw Areas, Waterbodies, Street Tree Inventory, 3D Massing, Neighbourhoods) and OpenStreetMap. The full ETL is reproducible from a single command; the methodology is fully spelled out in /methodology.
What this project is not
- It is not a definitive ranking of Toronto’s parks. Two parks doing different urban jobs cannot be reduced to a single number.
- It is not a livability index. We don’t score neighbourhoods or claim “best place to live”.
- It is not a safety model. We deliberately exclude policing or enforcement data. That conflation is harmful.
- It is not a replacement for visiting a park. The metrics catch built form, not programming, character, or community.
- It is not a finished product. New data sources and detector patterns get added as we encounter them.
Methodological caveats
- Measured ≠ truth. A high score does not mean a park is good. A low score does not mean it fails. Scores describe measurable conditions, not lived experience.
- OSM coverage skews downtown. Café and entrance density depends on volunteer mapping. Suburban edges are systematically under-detected.
- No real pedestrian counts. Activity is inferred from surrounding land use, not from observation. The model can’t see who actually shows up.
- Static snapshot. Seasonality, events, and time-of-day effects are not modelled. The current build is a single point in time.
- Confidence is layered. Each sub-score reports its own measured / partial / inferred level. The headline score doesn’t shrink low-confidence dimensions, but the UI flags them.
Different users value parks differently
A parent looking for a family-friendly weekend space will read the data differently than a runner mapping out their morning route, an event organiser scoping a 5-km circuit, an arborist tracking canopy loss, or a planner debating where to add a new park. Our model produces one consistent reading; it is one framework among many.
The two-axis Jacobs vs Wilderness chart is the single most important framing on the site. Toronto has lots of parks that are great on one axis and weak on the other; a balanced hybrid is rare. Knowing which axis you care about for a given purpose changes which parks are “good”.
Transparency
- Every score is explainable. The detail page shows the contribution of every dimension, the percentile rank, the inputs that produced it, and the source coverage.
- The ETL pipeline is reproducible end-to-end from raw open data. Every score traces back to source files and a documented transformation.
- The classifiers (typology, cluster, narrative) are rule-based and explainable. There is no opaque ML in the scoring or classification layers.
- Public feedback is collected per park (thumbs up / thumbs down + optional comment) and surfaced on /insights/contested. Where the model and the public disagree, we flag it.
- A timestamped snapshot of the final cache + manifest is produced with each build so any reported finding can be reproduced against the exact data that was on the page that day.
Project goals
To help Torontonians, planners, designers, and researchers understand how different kinds of parks function within the city’s urban fabric. Not to declare winners. Not to optimise. Not to recommend. To make the tradeoffs visible at the scale of the Toronto Park Catalogue, in plain language, with the data spelled out so anyone can disagree productively.
References & further reading
The intellectual lineage is documented at /references.
Built by
Colin Smillie
Founder & Developer
Colin builds tools that make Canadian public information easier to find and use. His work covers AI-powered municipal services, open data platforms, and civic-research projects like this one. The Toronto Parks Atlas grew out of a re-read of Jane Jacobs and a hunch that the slogans hadn’t held up to the way the city actually talks about its parks.
The Toronto Parks Atlas is an independent civic-research project. We are not affiliated with or endorsed by the City of Toronto. Data comes from City of Toronto Open Data, published under the Open Government Licence (Toronto), and from OpenStreetMap (ODbL).