Guides

Annual Speeds Data

Overview

Replica provides two speed-related datasets on an annual cadence: free-flow speed per network link, and a quarter-hourly speed profile per network link. Both datasets are produced with the same methodology.

Quality Metrics

Benchmarking

Comparison to New York City 2021 Traffic Camera Speed Data

We compare our quarter-hourly speed profile to New York City traffic camera speed data, averaged over the same collection period. This speed data is publicly available for download on the NYC Open Data Portal. Traffic camera speed data is analogous to our method, as cameras record the average speed between two points on a roadway via the elapsed time between them. The locations compared are shown on the map below.

Location of Comparison Points between NYC Traffic Camera data and Replica's Quarter-hourly Speed Profile

Below is a comparison between the time-averaged traffic camera data (“SPEED”) with the Replica quarter-hourly speed profile (“replica_speed”) on a section of the Trans-Manhattan Expressway (I-95), for a typical Tuesday.

Comparison between NYC Traffic Camera data and Replica's Quarter-hourly Speed Profile for one location

The histogram below shows the distribution of correlation coefficients computed between the Replica quarter-hourly speed profile and the traffic camera speeds time series on a per-location, per-day-of-week basis. The bulk of the two time series panels are strongly correlated, with a median correlation coefficient of 0.71.

Histogram of correlation coefficients between NYC Traffic Camera data and Replica's Quarter-hourly Speed Profile

Methodology

Replica's speed data is computed from a panel of time-stamped GPS traces, collected over a six-month period for each year. Speed values represent average speeds computed from the measured time to traverse the length of a network link, and the length of that segment of roadway. Thus, the speed values are the relevant measure of speed for constructing travel time tables, but are not the instantaneous speed one would measure from a radar-based sensor.

The average speeds data is aggregated to produce both the free-flow speeds per network link, and the quarter-hourly speed profile per network link tables. For both tables, the data are aggregated spatially at the level of OSM way. For ease of comparison with Replica's AADT and seasonal trip tables, we also associate each direction (indicated by “forward”) of each OSM way with more granular subdivisions of the street network.

Quarter-hourly speeds are provided for most major roadways during peak and mid-day hours. Speed data for a given quarter-hour is not reported when the observed vehicle count is low. This happens most frequently during overnight hours, when speeds are, presumably, at free-flow.