Global Nature Watch (GNW) Horizon · concept · workflow sketch
Point Horizon at your own data
Connect a dataset you already host somewhere else, then ask questions about it in chat. Nothing is uploaded and no copy of your data is stored.
Where this would sit in Horizon
Today the catalogue only holds GNW's own layers. The sketch below is what happens after this new entry point.
Different from a boundary upload — a boundary only tells Horizon where to analyze GNW's own data. Connecting a dataset brings your own attributes and values into the analysis itself.
The orange notes are things to think about as you review each section. The survey you will complete after reviewing asks related questions.
STEP 1 OF 5
Point to your data
Vector data only (shapes, lines, points)
STAC catalog
Google Earth Engine
ArcGIS Online
GeoNode
OGC web service (WFS)
Collection path or ID
https://your-geonode.org/geoserver/wfs?typeName=concession_boundaries
What data would you want to connect here? Where does the data live?
This connects to location data like points, lines, and boundaries (i.e., vector only) — not imagery or other pixel-based files. Does that leave out something you need?
STEP 2 OF 5
What do you want to know?
Summarize values in my area
Check overlap with my boundary
Measure change over time
Compare two time periods
Buffer-zone analysis
STEP 3 OF 5
Summarize values in my area
Detected from your data — check this looks right:
EPSG:4326
Geometry: Polygon (342 features)
Date field: issued_date
Status field: concession_status (active, expired, pending, revoked)
Which field should we summarize?
Select a field…
Same screen, when detection is incomplete
STEP 3 OF 5 · PARTIAL DETECTION
Summarize values in my area
We could only read part of this. Please confirm the rest:
EPSG:4326
Status field: concession_status (active, expired, pending, revoked)
Geometry type: not detected — please confirm
Select a field…
Date field: not detected — please select from your dataset's fields
Select a field…
STEP 4 OF 5
A few questions about your data
Generated from what Step 3 could not work out from the values alone.
concession_status: active, expired, pending, revoked — what does “pending” mean here?
From your data
e.g. Application filed but not yet granted.
issued_date: 1998–2024 — is this when the concession was granted, or when it takes effect?
From your data
e.g. Date of grant; effective date is 90 days later.
As of what date is this accurate?
2026-06-30
Is there anything this dataset leaves out that someone comparing it to GNW data should know?
Optional
e.g. Doesn't include informally held or unregistered concessions.
Add a note about terminology
STEP 5 OF 5
Ask your question in chat
your dataset
GNW data
queried live each time — never copied
How much tree cover loss has occurred within each concession type since 2001?
Ask
What are two or three questions you would actually want to ask that combine your own data with GNW data?
What you get
Your dataset appears on the Horizon map next to GNW layers, and the analysis is returned as a chart and a written answer.
Returning user behavior
What persists across sessions
The connection persists. Once a user points Horizon at a dataset, the connection (provider, path/collection ID) is remembered for that user going forward — they don't need to re-enter it each session.
The connected dataset appears in the user's data catalog — visible to that user only. It's listed alongside GNW's own datasets so it's easy to find and reuse across analyses, but it is not shared, published, or made discoverable to other users.
A pinned dashboard analysis persists but does not auto-recompute. A saved analysis using connected data does not automatically refresh; the user must manually refresh.
Pinned to a dashboard
What's stored on the dashboard is the provider, path, chosen analysis and parameters — not the data behind the chart.
What this does not do
No file upload, ever. You point to the data where it already lives. GNW never takes custody of a copy.
No map styling or analysis builders. Rendering, layer design and parameter tuning stay in the tools you already use. Horizon answers in text.
Only the reference is saved, never the data. Horizon remembers where your data lives — provider, path, chosen analysis — so you don't re-enter it. It never stores a copy, and no standing instance is created for you or your organisation.
Separate from boundary upload. A boundary only says where to look inside GNW's data. This says what to look at — your own attributes and values GNW doesn't have.
No accuracy guarantee. GNW hasn't verified this connected dataset and can't confirm the accuracy of results built on it.
What is your overall reaction to this concept? What questions do you have?