Workflow to create global mapbox tileset

Question

I am in process of creating a hiking map rendered from OSM data using Mapbox. It works quite fine already, however some data are missing in the Mapbox Streets V7 datasets, namely:

• bus stops
• power lines
• marked hiking/cycling routes

I am able to get all those data using Overpass Turbo for small regions and upload them as tilesets, however the largest area I am able to process this way is about 200x200 km for highly detailed areas like Europe.

Given the quite generous 50 GB limit for tilesets, I think I should be able to achieve a global (or at least continental) coverage with custom tilesets for those features. However I am not sure how to proceed. I have found an article which describes a workflow like this, however it does not show any implementation details:

Building a Custom Trail Map Using OSM and Mapbox

I think it could be possible to build some tool which would download data using Overpass API part by part. I understand the tools needs to avoid being too heavy on Overpass servers and I am fine with that, I can develop my queries using small region only and I am fine with downloading global data for a few weeks if needed. However I am bit concerned with issues about duplicate / missing data on the part boundaries. Is there any tool which would be able do handle the splitting and merging of the Overpass requests and responses (or at least merging the responses)? If not, is there any advice how to build such tool?

Answer 1

As suggested by mmd, it seems more practical to process planet or region files locally. I was surprised how are the tools performing the processing fast. The only operation which really took look for me was downloading the data.

Filtering / manipulating data using OsmComvert + OsmFilter (osmctools) or Osmium is quite easy, I think easier then trying to invent complicated Overpass QL querys.

My current workflow is:

• use osmconvert to add/modify tags and convert to o5m format
• use osmfilter to select just the data I want and convert to osm format
• use ogr2ogr to extract individual layers as GeoJSON
• use tippecanoe to merge the layers and process them into MBTiles

Currently I am processing about 3 GB of data (three European countries) and it takes about 10 minutes to complete all the steps above.