Facebook Commuting Zones

Step 1: Assign populations to home and work locations We assign populations to home and work locations based on information people provide and location services they have opted into (to learn see Location Privacy Basics).

Step 2: Calculate movement between locations We aggregate populations’ home and work locations to create a graph that measures movements between places over the previous few weeks.

Step 3: Identify clusters We use a community detection algorithm to identify clusters that represent commuting zones, and build our commuting zone shapes by dividing them into areas based on their centers.

Full details available here: https://dataforgood.facebook.com/dfg/docs/methodology-commuting-zones

Codebook: region - The continent which the Commuting Zone is located within.

fbcz_id - A unique identifier generated by combining the region name and fbcz_id_num value.

name - The name of the administrative region that most closely resembles the shape and location of the commuting zone. These are roughly equivalent to the 3rd administrative level, for example counties in the United States.

fbcz_id_num - A randomly generated unique number used to uniquely identify each Commuting Zone. These change with each version of Commuting Zones (but the name does not have to change).

cz_gen_ds - The date when the Commuting Zone geometry and metrics were generated. Provided in M/D/YY format.

win_population - Estimated population within the zone (calculated from the publicly available Facebook High Resolution Population Density Maps or WorldPop estimates). These population estimates are provided as counts per grid tile on the earth’s surface. We map each of these tiles to a commuting zone then aggregate population by taking the sum of all the tiles within the commuting zone polygon defined in the geometry column/field. We then winsorize the bottom and top 5% of commuting zones. This means that the population counts per Commuting Zone below the 5% percentile are replaced by the 5% percentile value; and above the 95% percentile are replaced by the 95% percentile value. All commuting zones have a population of at least 50.

win_roads_km - Estimated length of roads within the zone in kilometers (calculated from the publicly available OpenStreetMap and Facebook missing roads estimates). We calculate this value by taking the sum of the length of all roads within the commuting zone polygon defined in the geometry column/field. We then winsorize the bottom and top 5% of commuting zones. This means that the road length counts per Commuting Zone below the 5% percentile are replaced by the 5% percentile value; and above the 95% percentile are replaced by the 95% percentile value.

area - Area of commuting zone (in square kilometers). All commuting zones are larger than 1 km by 1 km.

country - The country where the majority of the Commuting Zone geometry is located within. We use the country boundaries provided by the GADM project.

geometry - The polygon shapes that define the boundaries of the commuting zone on the earth’s surface. These are a collection of latitude and longitude points on the earth's surface, that when connected by lines define a closed polygon (or multiple closed polygons) boundary. Note that this means that a Commuting Zone does not necessarily have to be a single polygon that is fully connected, and there can be “islands” of multiple shapes that are part of a single zone. Most geospatial software tools / libraries can easily recognize this POLYGON/MULTIPOLYGON format.

Sensitive geographies are removed from this dataset. A version with additional infrastructure statistics is available here: https://dataforgood.facebook.com/dfg/tools/commuting-zones

These zones are a set of boundary shapes built using aggregated estimates of home and work locations. Data used to build commuting zones is aggregated and de-identified.