What to filter out if making world map from OpenStreetMap data?

Question

I'm making a world pin map, high resolution, high detail map.

I've downloaded the planet-latest and played around with it. Before importing the file into a database I need to filter out nodes/objects so that it fits on my harddrive.

I am a bit lost as to what tags can I remove to still be able to make a good world map. I've been playing around with osmfilter and osmconvert.

Here's an example of how many tags are in the dataset.

  340712803 building
  195127703 source
  145411170 highway
   87781369 addr:housenumber
   81280839 addr:street
   68157172 name
   64141428 addr:city
   58872520 addr:postcode
   35372957 natural
   26512596 addr:country
   25334436 surface
   25106849 source:date
   24669992 landuse
   18837161 power
   16953011 waterway

I'm pretty sure I can remove building, but I have a hard time figuring out what is safe to remove.

Answer 1

It is not so easy to answer your question when we don't know exactly how your map should look like, or what is your detailed usecase.

But what about NOT asking what data shouls be removed, but asking yourself what data to KEEP before proceeding?

See the --keep parameter of osmfilter!

So why not ask yourself:

Do you need streets and highways? Same for railway infrastructure? you need landuse data? and waterbodies and coastlines? Boundary lines between states/countries?

And what rendering engine do you want to use?

Answer 2

All common tags are documented in the OSM Wiki.

• Do you want to show buildings in your map? If not, drop all building tags.
• You can safely drop the source tags since they are not useful for rendering.
• You will very likely need roads, i.e. highway tags.
• Do you want to render addresses? If not, drop addr:housenumber, addr:street, addr:city, addr:postcode, addr:country etc.
• You will very likely want to keep names (on roads, POIs, rivers, buildings, ...).
• The natural key is used for woods, grass, water, coastlines, ...
• You can drop surface except if you want to render the specific surface of roads in different styles.
• ...

It might also help to take a look at taginfo for determining the most common keys and most common values for a key, e.g. for natural=*.