Overture Maps

Last week Overture Maps released a new world-wide dataset https://overturemaps.org/overture-maps-foundation-releases-first-world-wide-open-map-dataset/ (2023-07-26). The first release of this dataset was in April 2023 (https://overturemaps.org/download/overture-april-alpha-release-notes/).

Overture Maps is an effort of Amazon, Meta, Microsoft, TomTom and Esri, so big companies but notably missing are Apple and Google.

At https://overturemaps.org/download/ we can get more information about the data. There are 4 themes: Places, Buildings, Transportation and Administrative boundaries. The bulk of the data is processed OpenStreetMap data, but other sources can be added.

The data is available under licenses like CDLA Permissive v 2.0 and ODbL. The data is stored in Parquet files in Amazon S3 and Microsoft Azure, the schema of the data we can find at https://docs.overturemaps.org/

At https://github.com/OvertureMaps/data/blob/main/README.md#data-location the download urls are shown:

image

To browse/download the data on Azure we can use Microsoft Azure Storage Explorer, it looks like:

image

Poking around the data, we can see it’s quite big (totally 200 GB): places is 8 GB and buildings is 110 GB.

Let’s download the first buildings file:

image

It’s a bit of cryptic name that ‘20230725_211555_00082_tpd52_00e93efa-24f4-4014-b65a-38c7d2854ab4’ but it should be a Parquet file. So let’s start DuckDB on the command line and inspect it:

image

So in this file there are 8 million buildings, they have a geometry (WKB as blob) and some other attributes. After ‘D copy buildings to buildings.parquet’ we have a new file buildings.parquet (1.4 GB).

We can analyze the geometries with the DuckDB Spatial functions, for example get the centroids of the buildings:

image

The spatial functions are a subset of the PostGIS spatial functions, see for more information https://duckdb.org/docs/extensions/spatial.html 

But can we visualize this file in QGIS? Yes, but first we have to convert it to a GeoParquet (https://geoparquet.org/) file. A GeoParquet file is a regular Parquet file, but with some extra metadata about the geometries. See https://bertt.wordpress.com/2022/12/20/geoparquet-geospatial-vector-data-using-apache-parquet/ for more detailed information about GeoParquet.

Tool ‘gpq’ (https://github.com/planetlabs/gpq) can be used to convert from Parquet to GeoParquet.

image

Note: directly reading of the Overture Parquet files in gpq does not work yet, see https://github.com/planetlabs/gpq/issues/57

After drag and drop the file ‘buildings_geo.parquet’ to QGIS it looks like:

image

So the buildings are scattered around, the building parquet files are not spatially ordered. To get all the buildings in an area we have to download more building Parquet files.

In some cities the attributes height and numfloors are filled in, for example in Boston

image

So in these places we can make 3D Maps, like https://github.com/shi-works/Overture-Maps-Data-for-GIS

image

Some other nice sites with processed data:

– All the places in a map https://bdon.github.io/overture-tiles/places.html

image

– Buildings GeoParquet files per country https://beta.source.coop/cholmes/overture/browse/geoparquet-country

image

Do you know any other sites with processed data? Let us know in the comments!

2 thoughts on “Overture Maps

  1. Hi Bert, interessant maar waarom niet gewoon OSM gebruiken en steunen? Wat ik zie is nog heel schamel en ook niet foutloos, zo zie ik Fort Vechten bij mij om de hoek ligggen ipv bij het fort zelf? Groet Victor

  2. Ha Viktor, dit is niet alleen OSM, maar ook POI’s van Facebook, gebouwen van Microsoft/Esri en regio’s van TomTom (en more to come). Mooi dat er zo meer data makkelijk beschikbaar wordt. Waar zie je dat fort Vechten, bij de Poi’s?

Leave a comment