Section 6 Technical Notes - "Congestion Condensed"
Data Sources
Just under 500 Automatic Traffic Counting ( ATC) sites are spread throughout the 10 areas being subjected to congestion monitoring. These sites collect traffic flow and speed information and store this in 15 minute bins, on a permanent basis - although they do periodically suffer damage due to highway maintenance works and occasionally from internal failures. The speed and flow data from these sites is the basis of all congestion monitoring work.
Note that some of these sites actually store data internally in vehicle-by-vehicle ( VBV) mode making it possible to reanalyze the data on a more precise basis to better identify the start and end of congestion, for example.
A series of Floating Vehicle surveys were also carried out in 2003 which collected speed, direction, location and time information at a frequency of once every second using GPS technology. The surveys were carried out during weekdays and Saturdays (in some cases) using between 4 and 6 vehicles each day, with a resulting total number of vehicle survey days of 344 (each day's survey between 7am and 7pm). This information was interrogated for time and speed data and used to produce calibration factors as explained below.
Since the congestion project was not initiated until early 2003 and all sites required either to be upgraded from an existing traffic-flow-only capability (ie no speed) and had to be installed from new, none of the sites have a complete set of data for 2003 with most sites coming on-stream between April 2003 and September 2003. The method summarized below takes into account these (and any other) gaps in data at specific sites in the process.
Methodology
A very brief explanation of the methodology: Speed = Distance / Time => Time = Distance / Speed.
If we restrict initial calculations to distances of 1 kilometre then Time = 1 / Speed
If we measure the Free Flow Speed ( FFS) - ie the speed at which vehicles will travel over a site when not restrained by other vehicles - then we can calculate congestion, as an increase in journey time over the length of route associated with that site in any single 15 minute time period as:
Lost time per vehicle kilometre ( LTVK) = 1 / V - 1 / FFS
(where "V" = the average speed of all vehicles in that time period)
If all sites were located within quite short sections of road and no priority junctions were involved then the above equation would be enough. However, even on the trunk road network in Scotland there are junctions where trunk route travelers must give way to other road users.
Consider Sherriffhall Junction on the A720 City Bypass. A simple traffic counting site located on the A720 to the east of that junction will pick up the congestion due to decreased speeds along that section of the route but will not pick up the congestion due to queuing at the junction nor decelerating on approach to the junction nor accelerating away from the junction.
A series of floating vehicle surveys was carried out, resulting in something of the order of 260 trips across each traffic counting site and the data used to calculate actual travel times between link ends (a link is the section of road associated with each site) as well as contemporary speed at the location of the sites. A regression analysis then produced a calibration factor "m" which was used to adjust the LTVK equation above as follows:
LTVK = 1 / mV - 1 / m FFS
Most other congestion indices calculated for a single 15 minute period for a single site are basic summations of LTVK. The daily congestion index is different in that it is an average over each day (weighted by flow per 15 minutes) of FFS/V. Journey Time Reliability is also different and is explained in the introduction to this report.
The relevant site-based congestion numbers are then summated, averaged, for all sites in a route.