Why does data licensing matter for geospatial products?

Licensing defines what you may legally do with source data: use it internally, redistribute it, sell derivatives, or publish it openly. Getting it wrong exposes you to takedowns, breach claims, and the need to rebuild a product from clean sources.

What is the difference between common open data licenses like CC-BY, ODbL, and OGL?

CC-BY requires attribution; CC-BY-SA and ODbL add share-alike (derivatives must use the same license); OGL is a permissive UK government license requiring attribution. OpenStreetMap uses ODbL, which can impose share-alike on produced databases.

Does combining datasets create share-alike or attribution obligations?

Yes. A product inherits the obligations of every source: you must satisfy all attribution requirements, and a single share-alike (ODbL/CC-BY-SA) source can force the whole derived database under share-alike terms unless the data is kept separable.

How do I track licensing and provenance across many sources?

Record license, source, version, date, and attribution text per dataset in a source register, embed it in metadata (ISO 19115 constraints fields), and propagate it through each processing step so the final product carries a complete, auditable rights trail.

Data Licensing for Geospatial Products

Short answer

Data licensing decides what you are legally allowed to do with the data inside a geospatial product — use it internally, redistribute it, sell derivatives, or publish it openly — and a product inherits the obligations of every source it touches. The recurring failures are not exotic: a share-alike source (OpenStreetMap's ODbL, or CC-BY-SA) silently dragging a commercial product under copyleft terms; missing attribution; or using "free to view" data in a way the license never permitted. Treat licensing as a data-model concern, captured per source and propagated through every processing step, exactly like CRS or schema — not as a legal afterthought bolted on before delivery.

A note on scope

This is practical guidance from building spatial products, not legal advice; licensing terms vary by jurisdiction and the database-rights regime that applies (the EU sui generis database right, for instance, has no direct US equivalent). For anything high-stakes, have counsel review the actual license texts. What follows is the engineering discipline that keeps you out of trouble in the first place.

The licenses you will actually meet

Permissive — attribution only

CC-BY 4.0 — use, modify, redistribute, even commercially, if you attribute the source as specified.
OGL (UK Open Government Licence) — very permissive; attribution required. Much UK public data (Ordnance Survey OpenData, etc.) uses it.
Public domain / CC0 — no conditions. Most US federal data (USGS, NOAA, NASA, USGS 3DEP, Landsat) is public domain because US government works generally are not copyrightable, though "courtesy" attribution is still good practice.

Share-alike / copyleft — the trap

ODbL (Open Database License) — used by OpenStreetMap. Attribution plus share-alike: a "Derivative Database" must be offered under ODbL, and a "Produced Work" (e.g. a rendered map) must carry attribution. The dangerous case is mixing OSM geometry into your own database — the produced database can fall under ODbL.
CC-BY-SA — derivatives must be licensed under the same terms. Combining CC-BY-SA data into a product you intended to keep proprietary forces the share-alike obligation onto the result.

Proprietary / restricted

Commercial imagery and basemaps (most high-resolution satellite, many commercial DEMs) come with explicit limits: no redistribution, no derivative resale, seat or area caps, sometimes "internal use only." Copernicus Sentinel and Landsat are notable open exceptions and are the backbone of many open products precisely because their terms are generous.

The two things that bite

1. Obligations are cumulative

Your product must satisfy the union of all source requirements. If you blend a CC-BY DEM, an ODbL road network, and a CC-BY-SA land-cover layer, you owe attribution to all three and you must reconcile two different share-alike regimes — which may be incompatible with each other and with a commercial release. The cleanest mitigation is keep share-alike sources separable: serve OSM as its own attributed layer rather than fusing its geometry into your proprietary database, so the share-alike obligation does not propagate into the rest of the product.

2. "Viewable" is not "reusable"

A WMS you can display in QGIS, a basemap visible in a web map, or data behind a public viewer is frequently licensed for viewing in context, not for download, extraction, redistribution, or training derivative datasets. Tracing features off a proprietary basemap to build your own layer is a classic violation. Always check the redistribution and derivative clauses, not just whether the data loads.

Tracking provenance so it stays auditable

You cannot honour licenses you cannot trace. Build a source register and propagate it:

Per-source record — for every dataset: name, publisher, URL, version, acquisition date, license (with a link to the exact text), required attribution string, and any restriction (no-redistribution, share-alike, area cap).
Embed in metadata — ISO 19115 has dedicated constraint fields (MD_LegalConstraints: use limitations, access constraints, and a license element). Put the license and attribution there so it travels inside the dataset, not in a forgotten spreadsheet.
Propagate through processing — when a step combines or derives data, carry forward the union of source licenses and attributions into the output's metadata. A reprojection, clip, or join does not erase the upstream obligations.
Surface attribution at the point of use — in a web product, render the combined attribution in the map credits/UI; for a downloadable dataset, ship a LICENSE/ATTRIBUTION file and the metadata.

Worked example: a flood-screening web product

Sources: USGS 3DEP DEM (public domain), Copernicus Sentinel-2 land cover (open, attribution requested), OSM building footprints (ODbL), and a commercial high-resolution basemap (proprietary, view-only).

The DEM and Sentinel derivatives can be freely fused and redistributed; add courtesy attribution.
OSM buildings stay a separate attributed layer, served from their own store, with the ODbL attribution shown — so the share-alike obligation does not reach across the rest of the database.
The commercial basemap is used only as a display tile layer under its license; no features are traced from it, and it is not redistributed or cached beyond what the terms allow.
The product's metadata and UI credits list all four sources with their required attribution strings; the source register records versions and license URLs for audit.

Result: a product you can ship and update without a rights cloud over it.

Common pitfalls and why they happen

Fusing ODbL/CC-BY-SA data into a proprietary database — share-alike then claims the whole produced database; the fix is separability.
Dropping attribution after processing — a clip or reproject "loses" the source note because licensing was never in the metadata, only in someone's memory.
Treating view access as reuse rights — tracing or extracting from a view-only basemap or WMS.
Assuming "open" means "no conditions" — open licenses still impose attribution and sometimes share-alike.
No version pinning — if a source's license or content changes, you cannot prove what terms applied when you built the product.
Mixing incompatible share-alike licenses — two copyleft sources whose terms cannot both be satisfied in one combined database.

QA and validation checklist

Every source has a register entry: license, version, date, URL, attribution string, restrictions.
License and attribution embedded in the product metadata (ISO 19115 constraints), not just a sidecar note.
Share-alike sources kept separable; their obligation does not propagate into proprietary parts.
Redistribution and derivative clauses checked for each source, not just viewability.
Combined attribution shown at the point of use (UI credits or LICENSE file).
Source versions pinned so applicable terms are provable for any product version.

Bathyl perspective

We treat licensing as part of the data architecture, recorded per source and carried through every transformation, so a spatial product can be shipped, updated, and resold without a rights question hanging over it. The discipline is mundane and it pays off exactly when it matters — when a client asks "can we publish this commercially?" and the answer is a documented yes, not a scramble to rebuild from clean sources.