Scooting along

Calgary is finishing the second year of a pilot of e-scooter share systems; one of the nice things is that some e-scooter data is available for public use on the City’s website.

I hope to do a couple of posts; the first is a high level look at the data, and then I want to look at a few questions to get a sense of how much scooters are being used for different purposes, like the first mile / last mile to transit.

With any dataset, there are pros and cons. The biggest pro here is it’s all scooter trips (or at least all trips over 30 seconds / 100 metres) that were made during the summer of 2019, so there’s no issue around sampling. The biggest challenge is the privacy protection measures. To protect the privacy of users, the trip data has been aggregated to the nearest hour, which shouldn’t affect any of my analysis.

Meet the grid

Again as part of privacy protection, there’s no information available to the public about the specific path any scooter took, although there is a map showing the most used street and pathway segments across all trips. However, the origins and destinations of each trip are available and have been geocoded into a hexagonal grid, with hexagons roughly 100m along each side, or containing roughly 30,000 sq m of land. The grids look like this:

The grid above is the City Hall area; the grid cells are roughly a block, but they don’t line up with any block in particular. Grid cell DU-104 contains the new Central Library, but it also has the back side of the Municipal Building, the eastern end of the City Hall LRT station, and the southern entrance to Bow Valley College. So they provide a broad sense of where the scooter started or ended the trip, but it’s difficult to say exactly where someone was going; which I suppose is the point of protecting privacy around origins and destinations.

Where the trips are

With this in mind, let’s look at where scooters are coming and going to. These figures show trip ends, which is the average of the number of trips starting and ending in each grid cell. (So each trip is counted half in the cell where it starts, and half in the cell where it ends.) There were over 480,000 trips made in the dataset. Here’s the city as a whole:

There’s broad usage across the entire city, but that can be pretty misleading — almost all of these cells have fewer than 10 trip ends in them out of nearly half a million. About 3/4 of the cells are in the under 10 dark purple category, and all those cells only combine to have about 2% of the trips. The figure below shows the distribution of cells and the share of trips.

The vast majority of trips – about 70% – begin and end in the light orange and yellow cells, which are only about 2% of the cells with scooter trips. Here’s the centre of the city; you can see where these concentrated areas of usage are; the slider allows you to focus on just the three top categories, cells with over 250 trips (over the entire summer pilot period, 250 trips is only 3 trips per day on average). These top three categories represent about 85% of the trips, and they cover a fairly small area.

The highest usage densities are on Stephen Ave, north to Eau Claire, and along 17th Ave, as well as in Kensington, Mission, Chinatown and the East Village. In general, use in the downtown and Beltline is high. Secondary concentrations include Inglewood, and Bridgeland; the most distant concentration is the Marda Loop area.

Origins and Destinations

The trip ends are interesting, but it’s also useful to look at where trips come from and go to. This can get pretty complex pretty quickly, so I split the cells into four main sectors; the CBD (green), the Beltline (blue), the “collar” communities (those directly bordering the CBD or Beltline) and the “outer” area, i.e. the rest of the city. The first figure shows where the sectors are (the hex grid doesn’t exactly match community boundaries), and the second shows the distribution of trips between these sectors.

So about 57% of all trips are within the CBD/Beltline area, and most of the remainder involve the collar area; only about 10% of trips start, end (or both) outside this fairly small inner city area. Generally, a fairly short distance set of trips. The collar-collar trips are generally fairly short as well; I split the “collar” into four quadrants and about 90% of the collar-collar trips are within one quadrant.

Times of Day

And finally, I want to close by looking at the temporal distribution of trips. The figure below shows the average hourly trip volume by the hour of the trip and the day of the week; the single busiest hour is 4-5 PM on Saturday, when on average there were about 600 trips being made.

Going from the quiet early morning (least quiet early Saturday and Sunday), there’s a clear mini peak between 8 and 9 AM on weekdays, during the morning commute. In general, though, scooter use peaks in the afternoon and evening; Sundays have a fairly early peak around 3, while weekdays have another mini peak around 4-5 PM during the evening commute, but it’s somewhat swamped in the higher general usage of scooters in the evening, with the demand dying down around 10 PM or so; later on Friday and Saturday, earlier on Sunday and Monday.

The time distribution suggests a little bit of commuting; perhaps in the 300 or 400 person range. (For comparison, the two hour AM peak cordon count of people entering the downtown had around 40,000 each by transit and by car, and around 10,000 by bikes and foot in 2019.) But the time distribution is primarily consistent with leisure use; going to socialize, to restaurants, and so on.

There’s so much to dig into, but more details will have to wait for another post!