Etienne Côme, @comeetie@mapstodon.space, FOSSDEM 2023

3/2/23

Researcher Gustave Eiffel University COSYS/GRETTIA

Stats, Data, Visualisation, Carto, Transports,…

R,JS,Python,Cpp,…

Going back in 2014 !

Massive Grided statistical data

- derived from tax sheets
- 200m x 200m pixels
- limit of the statistical secret (11 households)

open data but :

- heavy files with tricky file formats and projection problems
- only usable with pain
- on dedicated software mapinfo / arcgis / qgis

Webmapping stack developped around the OSM project

\(\Rightarrow\) would allow to renew the data diffusion/visualiation approaches

Not the only one to see this opportunity

Oliver O’Brien did a quite similar proposal in the Uk in 2015 with data-shine

At that time vector tiles were slowly comming

- http://mike.teczno.com/notes/postgreslessness-mapnik-vectiles.html
- https://observablehq.com/@d3/vector-tiles
- https://gist.github.com/wboykinm/7393674

\(\Rightarrow\) Massive advantage for statistical data

Tool-chain :

- R to process the data and export geojson vector tiles
- leaflet for raster tiles
- d3 hook to render “vectors tiles” on canvas
- invisible canvas tiles for interactivivity
- details matter (color scales/map background/labels)

The modifiable areal unit problem (MAUP) is a statistical bias that occurs when you aggregate point data.

see, https://gisgeography.com/maup-modifiable-areal-unit-problem/

The modifiable areal unit problem (MAUP) is a statistical bias that occurs when you aggregate point data.

\(\rightarrow\) try aggregating the raw data at different scales and link them with zoom levels

solve two problems :

- vector tiles sizes !! we need to keep them small
- MAUP !! we want to explore several aggregation scales

- INSEE
- Journalists, on specific topics eg. segregation and school
- Urbanists for territorial diagnosis
- Transportation researchers
- Grided data user
- Curious people …

Vector tiles have since become a mature technology

- MVT format specifications 1.0 and 2.0

and are no widely available

- postgis can now produce mapbox vector tiles
- new webmapping tools mapbox-gl -> MapLibre,…

A lot easier now :

- pypixgrid python script to export MVT tiles from postgis
- customizable multiscale aggregation
- mapbox-gl / mapLibre for webmapping

- FrancePixels 2019 new data from insee

- FrancePixels Bati custom aggreation of IGN BD Topo on side topics
- FrancePixels 2023 new data from insee + temporal evolution
- gridViz by @julgaf (eurostats)
- datashine successor
- INSEE project
- …

Questions ? Feedback