About
A search, not a claim
Chestly is an open-source geospatial research project. It combines satellite imagery, terrain data and documented observations of surviving American chestnut trees to rank locations that may hold trees nobody has recorded.
The project exists because the search space is the problem. Millions of hectares of eastern forest cannot be walked by the number of people who care about this species. Narrowing that ground to a ranked list of places worth visiting is a job software can genuinely help with — and it is a different job from identifying a tree, which software cannot do from orbit.
Everything here is built to be checked. Predictions carry the model version, feature version and imagery year that produced them. Methods are documented rather than described as proprietary. The code is open.
What Chestly can and cannot do
What it can do
- Rank millions of hectares of forest by how closely conditions resemble documented chestnut sites.
- Narrow a state to a short list of search areas a person can realistically visit.
- Show its working: every prediction carries the model version, features and imagery year behind it.
- Decline to search where it has not been validated, and say so.
What it cannot do
- Identify a tree species from orbit. No satellite in service can, at the resolution a single crown occupies.
- Tell you a chestnut is present. A candidate is a place worth checking, not a detection.
- Find blighted understory sprouts, which are invisible beneath a closed canopy.
- Replace the person who walks in and looks up.
Built by Harrison Vicic
An independent, open-source project. Researchers, conservationists and citizen scientists working on American chestnut restoration are welcome to take part — particularly anyone able to check a candidate in the field, which is the step no amount of imagery replaces.
Get involved
Help check a candidate in the field
Field volunteers investigate high-priority search areas and file structured reports — photographs, measurements and observations. You are never asked to decide the species yourself; that judgement happens later, in front of people qualified to make it.
Approval only makes you eligible to sign in and claim a candidate. Exact coordinates are released for a candidate you have actively claimed, one at a time.
Data sources
- iNaturalist
Community observations of Castanea dentata, used as labelled training examples.
- Copernicus Sentinel-2
Multispectral imagery at 10–20 m, revisited every few days, providing the seasonal signal.
- USGS 3DEP and The National Map
Elevation, slope and aspect, and the topographic basemap.
- USGS NLCD
Land cover, used to exclude ground that cannot hold a chestnut before any scoring happens.
- The American Chestnut Foundation
Restoration research and the standards this project defers to on identification.
Responsible disclosure of tree locations
Surviving American chestnuts are rare and vulnerable to collection, vandalism and well-meaning foot traffic. Chestly never publishes the exact position of an unverified candidate or a newly confirmed tree. Public maps show a generalised area; precise coordinates stay server-side and are shared with the researchers and landowners who need them.