How can I download all hospitals in Africa from OpenStreetMap using Python and the Overpass API?

Question

I need a shapefile of every hospital in continental Africa from OSM. At this point, I've only played with downloading extracts from other services like geofabrik, I haven't actually pulled any data down myself. While I intend on learning how to use the Overpass API, I need a quick turnaround for this project. How can I use Python and the Overpass API to download a dataset of Africa's hospitals?

Answer 1

A spectacularly quick and short crash course:

• Download an.osm.pbf Africa extract from Geofabrik
• Use osmosis to filter it to keep only certain tags
• Convert the filtered data to a shapefile with ogr2ogr

The example below uses Congo only because it's smaller than all of Africa, and you may need to also add eg healthcare=* as a filter:

$ wget https://download.geofabrik.de/africa/congo-democratic-republic-latest.osm.pbf
(Progress...)
$ osmosis --read-pbf congo-democratic-republic-latest.osm.pbf --log-progress --tf accept-ways amenity=hospital,clinic,doctors,dentist --used-node --write-pbf congolese-hospitals.osm.pbf
INFO: Osmosis Version 0.48.2
INFO: Preparing pipeline.
INFO: Launching pipeline execution.
INFO: Pipeline executing, waiting for completion.
INFO: Processing Node 2096625632, 419655.26894621074 objects/second.
INFO: Processing Node 6901615718, 515231.3074770092 objects/second.
INFO: Processing Way 661735764, 877193.7612477505 objects/second.
INFO: Processing completion steps.
INFO: Completion steps took 13.833 seconds.
INFO: Processing complete.
INFO: Pipeline complete.
INFO: Total execution time: 90156 milliseconds.
$ ogr2ogr -f "ESRI Shapefile" congolese-hospitals ./congolese-hospitals.osm.pbf
0...10...20...30...40...50...60...70...80...90...100 - done.
Warning 6: Normalized/laundered field name: 'admin_level' to 'admin_leve'
ERROR 6: Geometry type of `Geometry Collection' not supported in shapefiles.  Type can be overridden with a layer creation option of SHPT=POINT/ARC/POLYGON/MULTIPOINT/POINTZ/ARCZ/POLYGONZ/MULTIPOINTZ/MULTIPATCH.
$

As you can see from the warnings and errors, shapefiles are slightly legacy; if GeoPackages are an option,

$ ogr2ogr -f GPKG congolese-hospitals.gpkg ./congolese-hospitals.osm.pbf
0...10...20...30...40...50...60...70...80...90...100 - done.

For proof, here's a part of the result being browsed in JOSM: