Waterbody reduction

Inland waterbodies in OpenStreetMap are mapped with a variety of tags producing a fairly large volume of data - more then 5GB of data in usual GIS formats for the polygon data alone. Since there is no objective measure of importance for these different features there is no easy way to filter from this large volume of data only those features that are relevant for rendering maps at coarse scales.

Therefore most OpenStreetMap based maps render water at the low zoom levels either from other data sources or using questionable filtering techniques that often lead to low quality results.

The reduced waterbody data we offer here is generated from the full set of waterbody data in OpenStreetMap and reduces this in volume without subjective decisions that affect the rendering results in undesirable ways. This is done by rasterizing the data at high resolution and generating a supersampled low resolution raster mask from it. This raster mask is made available as raster images and converted back to a very compact polygon representation which is made available as well.

This approach comes with a number of limitations that need to be kept in mind when using this data.

  • The files are only suited for rendering in the resolution they are produced for. Rendering either in a higher or lower resolution will produce suboptimal results.
  • The polygon version does not have any topological similarity with the original data as you might expect. The rivers for example do not form a continuous geometry.
  • The styling you can use with these files is somewhat limited. You cannot render the polygons with an outline for example. Using fill patterns or edge color gradients should however work.
  • The process does not perform any geometric generalization.

Technically the reduction process is implemented with Osmium and GDAL. The polygon representation is produced using a custom vectorization tool.

A more detailed background text provides additional information.

Rendering example

Here a rendering example derived from the OpenStreetMap standard style.