Alpha version

Automated data acquisition

1. Download OSM data from geofabrik.de for a given country, i.e. roads, paths, and transport-related points of interest (train & bus stations).
2. Download Land Cover data from Global Land Cover. NB: Data may be downloaded per tile (e.g. 13 tiles are needed to cover Africa).
3. Download topographic data from SRTM. NB: The python module elevation can be used to automatically download SRTM data for a given area of interest.
4. Download water data from Global Surface Water Explorer. A script is provided by the authors for automatic downloading (see here).
5. Get health facilities from Bluesquare.

Preprocessing of OSM data

1. Get the snapshot of OSM data corresponding to the period of interest.
2. Extract roads & other transport networks with all the relevant attributes.
3. Assign a travel speed for each road segment based on their attributes and--if necessary--on seasonnal, topographic or land cover information.
4. Convert geometries to a raster file (50m, 100m?) with cell values equal to travel speed.

Preprocessing of land cover data

1. Clip & reproject input land cover data.
2. Reclassify land cover legend into categories that are relevant to travel speed.
3. Assign a travel speed to each land cover category.
4. Convert to a raster file with cell values equal to travel speed.

Preprocessing of water data

1. Clip & reproject input water data.
2. Get the snapshot corresponding to the period of interest.
3. Reclassify legend into categories that are relevant to travel speed.
4. Assign a travel speed to each category.

Preprocessing of topography

1. Clip & reproject input elevation data.
2. Compute slope and aspect from digital elevation model.

Preprocessing of population data

1. Clip & reproject input population data.

Modeling

1. Create a final raster with cell values corresponding to travel speed.
2. Compute the friction raster.
3. Least-cost modeling.

Spatial metrics

1. Travel time to nearest health facility.
2. ...depending on the health facility and the services they provide.
3. ...depending on the transport mode (car, pedestrian, etc.)

Non-spatial metrics

1. Population without access to any health facility in less than m minutes.
2. ...depending on the population characteristics and the services provided by the health facilities.

Notes

• Some input variables are time-sensitive: transport network (was the road segment available in the OSM database at a given date?), water (seasonnality is taken into account in the Global Surface Water product). Furthermore, travel speeds in a given land cover may also be affected by seasonnal variations.

Pending decisions

• Spatial resolution: 50m or 100m ?
• How do we integrate the temporal dimension ?
  • By building one model for each month ?
  • By building one model for each season ?
  • By building a model for a given date ?
• Metrics & visualization