LICENSE

The files in this directory are copies and adaptations of datasets published
by Data for Good at Meta (now AI for Good at Meta) on the Humanitarian Data
Exchange (HDX):

  https://data.humdata.org/dataset/movement-distribution
  https://data.humdata.org/dataset/movement-range-maps
  https://data.humdata.org/dataset/commuting-zones
  https://data.humdata.org/dataset/facebook-business-activity-trends-during-crisis

HDX lists each of these datasets under the Creative Commons Attribution
International licence (CC BY, license_id cc-by); they are used here under
CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Each dataset's
licence was read from the HDX API when it was fetched and is recorded in
manifest.json; a dataset whose licence is not cc-by is not mirrored.

Credit:

  Data for Good at Meta, Movement Distribution, Humanitarian Data Exchange (https://data.humdata.org/dataset/movement-distribution)
  Data for Good at Meta, Movement Range Maps, Humanitarian Data Exchange (https://data.humdata.org/dataset/movement-range-maps)
  Data for Good at Meta, Commuting Zones, Humanitarian Data Exchange (https://data.humdata.org/dataset/commuting-zones)
  Data for Good at Meta, Business Activity Trends during Crisis, Humanitarian Data Exchange (https://data.humdata.org/dataset/facebook-business-activity-trends-during-crisis)

The files under each <dataset>/original/ directory are the files as
downloaded from HDX, unchanged except for their names (a tidy name with an
extension; the name on HDX is in manifest.json). Files that HDX has since
removed are kept.

The Parquet files were changed from the originals (2026-09-29):

  Movement Distribution (movement-distribution/):
    - The CSVs were combined into one Parquet file per year, with a
      source_file column naming the original file.
    - Where the same gadm_id, polygon_level, distance category and ds is in
      more than one CSV (the files overlap on the days at their ends), only
      the row from the file HDX changed last was kept; the counts are in
      manifest.json.
    - polygon_level became an integer, distance_category_ping_fraction a
      double and ds a date.

  Movement Range Maps (movement-range-maps/):
    - The TSV inside each zip file was read (README.txt and the __MACOSX
      entries were left out) and the rows were split into one Parquet file
      per year, with a source_file column naming the TSV.
    - ds became a date and the two measures doubles; polygon_name keeps the
      string NA of the source.

  Commuting Zones (commuting-zones/):
    - Converted to GeoParquet: a geometry column (WGS 84) was made from the
      WKT in the geography column, which is kept. The WKT of some zones is
      cut off in the source and cannot be read; their geometry is NULL and
      geometry_truncated is true.
    - fbcz_id_num became an integer and win_population, win_roads_km and
      area doubles.

  Business Activity Trends during Crisis (facebook-business-activity-trends-during-crisis/):
    - The five CSVs were combined into one Parquet file, with columns that
      only some files have filled with NULL, and a crisis column (from the
      file name) and a source_file column added.
    - activity_quantile, latitude and longitude became doubles and ds a
      date.

In every Parquet file, no value was changed: text stays as in the source,
numbers and dates were only given a type. Rows were sorted. The row count and
the sum of every number column were checked against the originals.

These files are provided as is, without warranty of any kind. This server is
not affiliated with Meta or HDX.