We just shipped a significant update to the data pipeline behind our Interactive Basemaps. The change fixes a class of visual inconsistencies at low- and mid-zoom levels, such as missing rivers, shifting coastlines, islands that suddenly appear, and major water features that pop into detail abruptly rather than transitioning smoothly. The root cause was mixing two different data sources at different zoom levels. The fix is a unified OSM-derived dataset across all zoom levels, sourced from Overture Maps.

Here's what was happening and what we changed.

Why Water Features Were Disappearing at Mid-Zoom Levels

Our basemaps were pulling from two different datasets, depending on the zoom level. At lower zooms, we used Natural Earth, a well-known dataset that efficiently represents the most important details for broad-scale cartography. At higher zooms, we switched to OSM-derived data. These two datasets don't agree on everything: coastlines have vastly different levels of detail and features that exist in one don't always have equivalents in the other at the same zoom thresholds.

The handoff point between them was the source of most of the visual inconsistencies. A customer noticed a clear example: The St. Lawrence River disappeared entirely between zooms 6 and 8. OSM represents it as a water polygon, while the way we processed Natural Earth's simplified data at those zoom levels didn't carry the equivalent coverage. The result is a river-shaped hole filled with land. But that was one symptom of a broader pattern. Coastlines were shifting shape mid-zoom, small islands were appearing or vanishing, and water features were popping in suddenly rather than transitioning smoothly.

afterbefore
BEFORE
AFTER

What Changed: A Single Data Source Across All Zoom Levels

We've unified our basemaps around a single OSM-derived dataset across all zoom levels. Specifically, we're using releases from Overture Maps, which apply additional QA on top of OSM to ensure major water bodies are complete and consistent, something we had occasionally struggled with when building directly from raw OSM data.

Using the same source throughout eliminates the cross-dataset seam. No more coastline jumps. No more islands appearing from nowhere. No more rivers losing their shape at mid-zoom. It also means features that were simply absent at lower zoom levels. Rivers, estuaries, and reservoirs are now present and correct from the moment they're geographically relevant, rather than popping in abruptly as you zoom past the old data handoff point.

We also updated our geometry simplification algorithm and parameters. The new approach carries more detail through mid-zoom levels and produces smoother transitions as you zoom in and out with less of the sudden snap from one level of detail to the next.

The St. Lawrence is back where it belongs: see it here.

What This Means for Your Application

If your application renders maps at mid-zoom levels, your users are now seeing more accurate, more consistent water rendering. Specifically:

  • Rivers, estuaries, and reservoirs now appear at the zoom levels where they're geographically relevant
  • Coastlines have smoother transitions, with more detail present at low and mid zoom levels
  • Islands and water bodies render consistently from zoom to zoom
  • Transitions between zoom levels are smoother across all water features

These are the kinds of artifacts that are easy to miss in development and easy to notice in production. If you ever spot anything on the map that still looks off, let us know at support@stadiamaps.com. Reports like these are how this fix started.