Urban Tree Reporting
Share a filtered city map
Map links use the hash route, for example:
https://greenmtnboy.github.io/tree_reporting/#/?city=USNYC&mode=explore&filter.species=Platanus+x+hispanica&colorBy=dbh.
Supported query parameters:
| Parameter | Meaning |
|---|
city | Published city code, such as USNYC or USSFO. |
mode=explore | Open the city map directly, including on mobile. A recognized nonempty filter also opens exploration when mode is omitted. |
filter.species | Exact published scientific name, including genus-only inventory entries. |
filter.genus | First word of the published scientific name; includes all matching species and genus-only entries. |
colorBy | dbh for trunk diameter in inches, or tree_form for the normal category colors. Filtered links default to DBH. |
Encode values with URLSearchParams; a slug or common name is not a species key.
Repeat a filter key for alternatives (OR); different filter fields combine with
AND. Unknown filter keys and empty values are ignored. An unmatched name shows
an empty map. URL filters intersect with chat filters and survive city changes,
reloads and browser history. Chat color overrides take precedence until a color
is selected in the legend. “Show all trees” clears URL filters and URL coloring;
any separate chat filter still applies. DBH's gray category includes missing,
nonfinite and nonpositive measurements, without substituting a rendered size.
Deploy this map support before publishing arborary.world's species links.
An interactive browser-based map of city trees. Explore hundreds of thousands of street trees by location, species, and ecological attributes.
Urban forests are a key part of modern urban design - they can reduce city heat, absorb runoff, improve air quality, and enhance the mental health of residents. Read more here.
Plus they look nice! Charismatic megaplants.
There are several projects that collect open tree data; this one aims to be more exploratory and fun. Raw data is available in parquet, and of course also from the original sources.
Currently has data from the following cities:
- San Francisco, CA, United States
- New York City, NY, United States
- Boston, MA, United States
- Burlington, VT, United States
- Washington, DC, United States
- Los Angeles, CA, United States
- Tempe, AZ, United States
- Denver, CO, United States
- Stoughton, WI, United States
- Vancouver, Canada
- Calgary, AB, Canada
- Edmonton, AB, Canada
- Winnipeg, MB, Canada
- Toronto, ON, Canada
- Montreal, QC, Canada
Live page
What it does
Shows you the trees of a city on a map. Typically these will be a subset of all trees; private land is often excluded. London, for example, has about a million trees in their dataset and estimates over 4 million total in the city.
Map Page
Pulls in trees onto a navigable map.
Clicking a tree opens a popup with its common name, species, planting date, trunk diameter, and enriched species data (native status, evergreen status, mature height, bloom season, wildlife value, fire risk, etc.).
For supported cities you'll also get a landmark list with a search filter; clicking one flies the map camera to that location.
A chat panel to use with LLMs (bring your own API key or use a test key) lets you interact with the map in natural language - ask questions, fly around, or update what's shown.
Data sources
Tree inventories
Every tree row carries a data_source column identifying which of the above it came from.
Landmarks
Species enrichment
Per-species attributes (native status, evergreen, mature height, canopy spread, growth rate, lifespan, drought tolerance, bloom season, wildlife value, fire risk, tree category, map icon) are pulled in from a combination of the following services:
- Wikipedia REST API and MediaWiki API
- Kew's World Checklist of Vascular Plants (WCVP, the names and distributions behind POWO; Govaerts R (ed.), Royal Botanic Gardens, Kew, CC BY) - sftp.kew.org/pub/data-repositories/WCVP/
- GBIF species APIs - api.gbif.org
- SelecTree (Cal Poly UFEI) APIs - selectree.calpoly.edu
:::info
Tree info has been automatically extracted from the available APIs per species label and may be inaccurate - corrections very welcome, especially by people that know trees. Don't cite this in your paper!
:::
Tree-level predictions
tree_predictions_v{n}.parquet carries a predicted crown width per tree, from
a genus-level stem-diameter power law fitted on the
Tallo database (Jucker et al. 2022,
Global Change Biology, doi:10.1111/gcb.16302;
CC BY 4.0). Tallo is forest-plot data, so the crowns run narrower than an
open-grown street tree's; the parquet is a model-curation input today and
does not drive the map. Refit with data/raw/tools/crown_allometry_fit.py --write.
A tree with a planting date and no measured diameter gets one predicted from
its age, by a genus-level power law fitted on the map's own dated, measured
trees on every build, and the crown model is applied to that;
crown_dbh_source says which diameter a row's crown came from.
Species images
Tree species photos are sourced from the iNaturalist API. Images are licensed by their original photographers under Creative Commons licenses. iNaturalist is a joint initiative of the California Academy of Sciences and the National Geographic Society.
Basemap
CARTO Dark Matter vector tile style.
Tech stack
If you're someone who cares:
| Layer | Library / Service |
|---|
| Framework | Vue 3 (TypeScript) |
| Build | Vite 6 |
| Map | MapLibre GL 4 |
| In-browser SQL | DuckDB WASM |
| State | Pinia |
| Routing | Vue Router 4 |
| Preprocessing/data access | Trilogy |
Data pipeline dependencies: Python 3.13, PyArrow, Pytrilogy, DuckDB, Pillow, instructor, google-genai. The fun stuff!
Repository
https://github.com/greenmtnboy/tree_reporting
Similar
Urban Forestry
Dev
City Additions
Each city is an independent ingest pipeline: its own freshness probe, its own
model, its own OSM extract, and its own three entries in data/trilogy.toml
that schedule them. Full runbook in EXTENDING.md; the short
version is
- ingest watermark script (
{city}_update_time.py)
- data ingest script for trees (
{city}_tree_info.py)
- optional - landmark script or empty CSV
- trilogy model (
{city}_tree_info.preql) + its import in raw/tree_info.preql
and its stub in raw/full_tree_publish.preql
- OSM extract model (
osm_staging/{code}_osm_staging.preql; the row script
osm_staging/osm_rows.py is shared and selects the city from a where clause)
- three
[[cloud.job]] entries in data/trilogy.toml: city-{code},
osm-{code}, and a landmarks-{code} if the landmarks are a curated CSV
- update
README.md links/attribution and src/src/data/sourceCatalog.ts so the info page stays in sync
data/raw/tests/test_cloud_jobs.py fails if a city is missing a job, so a
half-added city is a red test rather than a parquet that quietly never
rebuilds.
Update Data
The scheduled refresh runs on trilogy-cloud — data/trilogy.toml declares one
job per city plus a daily core, and .github/workflows/cloud-sync.yml deploys
them on merge to main. Nothing needs to run locally for data to stay current.
To rebuild one city by hand (GCS writes; bucket/locations would need to be
parameterized for anyone else):
cd data && trilogy refresh raw/ussfo/sf_tree_info.preql
or fire the deployed job, which needs no local Google credential at all:
trilogy cloud jobs run urban-tree-city-ussfo --wait
How often each portal actually publishes, measured rather than assumed:
cd data/raw && uv run ./tools/portal_cadence.py --record
