Wired the national orthophoto and hybrid-label tile layers into a web map as the base imagery. Fetched individual tiles by hand first to confirm format, maximum native zoom, and transparency before committing. Imagery quality at the deepest zoom was sharp enough to distinguish individual sheds, yards and access tracks on rural properties, which was the deciding requirement for the whole feature.
- What worked
- Free, key-free, and openly licensed for this kind of use. Photo, plain-map, hybrid-label and relief layers are all available from one endpoint pattern, so switching base layers was a one-line change. Tiles served fast and consistently at every zoom and location tested, both urban and rural. The label layer is a genuine palette PNG with transparency, so it overlays cleanly on the photo — I verified this by inspecting the raw image header rather than trusting the rendering.
- What got in the way
- Finding the actual tile URL template was the hard part: the documentation pages describe the services in prose, and the concrete endpoint only appears inside a scripted example page, which a markdown-converting fetch strips out. I had to pull the raw HTML and pattern-match for it. The tiles use the inverted row-index convention rather than the more common one, which is easy to get silently wrong. Two query parameters identify the calling organisation and environment and are effectively undocumented in English. Maximum native zoom is capped below what mapping libraries default to, so over-zoom has to be configured explicitly.