Rail transit and population density

Comparing and ranking 250+ cities around the world

Aniket Kali & Jeff Allen | December 2024

Code used to generate final metrics, maps, and otherwise analyze data, is available under analysis/.

Website code is available under src/.

Details about data sources are available on the webpage, and embedded links are visible at src/routes/+page.svelte.

Re-running the analysis, a how-to!

OSM data is updated regularly by users, so we want to make sure that we include updated data, especially for cities which are lacking in data.

You can do this in a few simple steps.

Setup

Ensure that you have the following data downloaded and named appropriately. All data should be under the folder ./data, stored under the root directory.

• ./data/ne_10m_populated_places/: Natural Earth
• ./data/GlobPOP_Count_30arc_2022_I32.tiff: GlobPOP, using the 2022 population count data
• ./data/src_water/v107/: World Water Bodies
• ./data/src_water/water_simp.gpkg: Apply the Douglas-Peucker simplification algorithm in QGIS to the World Water Bodies layer (or something else that reduces file size)
• The hillshade topography layer is sourced from ESRI

Create these directories as well, if they don't exist already, which are used for storing intermediate stages of data. Inside ./data/, you can run: mkdir city_pop_tiles city_water osm_data

Analysis

Next, run these programs.

1. get_city_list.py: Selects our subset of cities and relevant fields as city_list.gpkg (<1m).
2. query_osm.py: Obtains rail and station data in GeoJSON format (~30m).
3. get_city_water_poly.py: Creates GPKG files containing the water layer relevant to each city (~5m).
4. get_city_metrics.py: Computes the metrics for all cities as city_metrics.gpkg and city_metrics.json (~5h). This can be done in 2 parts if needed:
   1. Comment out compute_metrics, run once (~2.5h).
   2. Comment in compute_metrics, comment out get_city_pop_tiles, run again (~2.5h).
5. render_maps.ipynb: Compiles layers and generates graphs for each city (~30m).