CHAPTER FOUR INDICATORS AND MEASUREMENT OF CONGESTION
4.1 At the practical level of measuring congestion, more concrete indicators are needed. A wide number have been developed - some in the UK context but many in the USA, although literature suggests that only a small number form the basis for regular monitoring of the network. A summary of the approaches used is given here.
4.2 As part of a report on the role of a national road traffic reduction target, DfT (2000) produced summaries of traffic congestion alongside a number of traffic related impacts for England (such as pollutants, safety and social impacts). It conceded that whilst 'a number of transport commentators have attempted to estimate congestion, using a variety of definitions, an ideal measure has yet to be identified'. As input to that report, The Commission for Integrated Transport ( CfIT) advised a measure based on:
The total amount of delay encountered, calculated across all traffic from the difference between the actual speed encountered and free flow speed
4.3 This forms the basis for the National Transport Model forecasts ( DfT, 2003), which are then key inputs to the FORGE Road Capacity and Costs model ( DfT, 2005). In fact an alternative measure was used in the report by DfT (2000) which divides this estimate of total delay by the volume of traffic to give the average amount of delay encountered by a vehicle travelling one kilometre.
Average delay by a vehicle travelling one kilometre = total delay to travel one kilometre/volume of traffic
Where total delay = actual speed - free flow speed (for all vehicles)
4.4 This average delay calculation is incorporated in the Transport Model for Scotland congestion mapping process; output from which forms the basis for the analysis presented in chapter 3. The second measure was believed to be advantageous in providing a better picture of how changing traffic levels and different policy packages can affect time lost to congestion. A detailed illustration of the use of this measure, with assumptions and reference input data is given in DfT (2000b), where figures on road traffic congestion are produced by road class, time of day and geographical location for England in 2000. Neither measure, however, gives an indication of the variability in time taken for a specific journey, or the relative importance of delays to different types of journey. It should also be noted that delays are measured purely in terms of vehicle journey time and no allowances are made for differences in occupancy rates, values of time, or for additional factors such as additional operating or environmental impacts that congestion can generate.
4.5 Simple measures relating to speed are also used to indicate congestion, particularly for a motorway environment. A current example would be the M42 Active Traffic Management ( ATM) scheme (Grant-Muller, 2005) where eight separate indicators have been identified to demonstrate the impacts of ATM in changing levels of congestion, as shown in Table 4.1. Other work also advocates simple speed related measures, for example Dijker et al, 1998, who proposed that traffic is considered in congested state when the traffic speed is below 50km/hr. The different indicators in Table 5.1 are relatively straight forward measures individually, but intended to give a more comprehensive picture of different aspects of congestion when taken together. Although simple to calculate, the data requirements to produce all 8 indicators are substantial and involve continuous loop monitoring of the area. As loops do not provide actual journey times (rather inferred journey times from speed), additional journey time data would be preferably produced, either through ITIS, ANPR matching or surveys. It is beyond the scope of this work to elaborate on reliability of data sources, but it should be noted that ITIS, ANPR and surveys also have inaccuracies in reflecting the state of the system. Experience has shown the use of loops for mean journey time may be adequate, but using these to produce estimates of variability of journey times may be less satisfactory, with less correspondence between loop based data and other data sources on this indicator. Whilst loop based data generally supports speed based indicators, the accuracy of loop based data at low speeds (less than 25 mph) diminishes, bringing into question the ability to use this data source to generate data for the 25 mph threshold. In addition, where congestion is a result of incidents or unexpected phenomena, the algorithm to convert loop data into journey times performs less well.
Table 4.1 - Congestion indicators for the M42 ATM project
1. Mean Journey Times
Mean journey time on a link-by-link basis, for specified time periods These to be combined into meaningful journeys, e.g. full ATM section, by direction.
2. Variability of Journey Times
Standard deviation (variance) in journey times on a link-by-link basis, and on a route basis:
- within-day variability
- between-day variability
Total number of vehicles per time interval that pass a point on the carriageway
4. Total Time Speed Less Than 25mph and 50 mph
Total time during which the average speed of vehicles drops below 25/50mph, per pre-defined time interval and per section (between junctions)
5. Number of Occurrences Speed is Less Than 25mph and 50 mph
Number of vehicles with average speed below 25/50mph, per pre-defined time interval and per section (between junctions)
6. Queue Lengths
Four types of queue to be measured,
- queues due to flow breakdown
- queues at exit slip roads
- queues on on-slips
- queues to join the ATM section
Queuing traffic is defined as a platoon of vehicles whose speed does not rise above 25mph.
7. Speed differential between lanes
Difference in mean speeds between each of the lanes per section, plus difference in extremes in distribution
8. Delay per hour/day
Measure of delay per hour/day on the ATM stretch, where delay is reflected through difference between free flow and actual journey time.
Notes to table
Source: adapted from Grant-Muller (2005)
4.6 The congestion reference flow (Highways Agency, 1997) gives a quantified measure of congestion for a link as follows (junctions must be considered separately).
CRF = CAPACITY * NL * Wf * 100/PkF * 100/PkD * AADT/ AAWT
where CAPACITY is the maximum hourly lane throughput
NL is the Number of Lanes per direction;
Wf is a Width Factor
PkF is the proportion (percentage) of the total daily flow (2-way) that occurs in the peak hour;
PkD is the directional split (percentage) of the peak hour flow;
AADT is the Annual Average Daily Traffic flow on the link;
AAWT is the Annual Average Weekday Traffic flow on the link.
4.7 Suggested values that may be used in the calculation are given within Highways Agency, 1997. Links of the same standard will have different CRF values according to factors such as the proportion of heavy vehicles, the peak to daily ratio, the peak hour directional split and the weekday/weekly flow ratio.
4.8 The level of Service indicator ( LOS) is one of the basic congestion measures applied widely in the USA and which has also been proposed by the Scottish Office (1998). It uses a scale running from A to F to describe operational conditions on a route or section of route taking into account speed, travel time, manoeuvrability, disruption to flows, comfort, convenience and safety. An 'A' rating represents the highest quality of service with free-flow conditions and users travelling at their desired speed. On single carriageways, passing demand is significantly below passing capacity and no platoons of three or more vehicles occur. On dual carriageways and motorways, minor disruptions to flow are easily absorbed without changes in speed. At the other end of the scale, an 'F' rating represents the worst quality of service with heavily congested flows and traffic demand exceeding capacity. Passing is virtually impossible on single carriageways and, on dual carriageways and motorways, long queues form which are subject to stop/start conditions.
4.9 Summary indices can be used to give congestion measures for a wider area rather than particular links and the desirability of these will depend upon the end use of the measure. One example is that given by Leonard (1993), who outlines a travel time based Congestion Index for comparative use in urban areas:
Where CI = Congestion Index
t i = free flow travel time
d i = excess travel time
4.10 This can be applied for all vehicle journeys or for single links of corridors. Where links are summed separately, it is necessary to apply a flow weighting:
Where CI = Congestion Index
t a = free flow travel time on link a
d a = excess travel time on link a
f a = flow along link a
4.11 The choice of a summary index or more specific link/junction based measures depends upon the end use of the data. Where the objective is to identify or monitor particular points in the network - for example for the purposes of monitoring congestion problem sites - an index will lose the desirable granularity in the information. This may be the case where the intention is to provide information to the traveller to advise journey planning for example. Where the objective is to assess costs and benefits of a particular scheme or policy, a wider indicator of congestion (or series or indicators, as is the case with the M42 ATM) would provide better information.
4.12 A number of indicators have been developed and are commonly applied in the USA and these are summarized in Table 3.1 below. In addition to those reported here, a wide tranche of literature on incident detection algorithms exist, many of which involve heavy instrumentation of the highway and frequently a Neural Network based analysis. These are not discussed further here as they lie outside the scope of the work, but see for example Wang et al, 2005.
Table 4.2 - Congestion Indicators adopted in practice within the USA
Roadway Level Of Service ( LOS)
Intensity of congestion delays on a particular roadway or at an intersection, rated from A (uncongested) to F (extremely congested).
Travel Time Rate
The ratio of peak period to free-flow travel times, considering only reoccurring delays (normal congestion delays).
Travel Time Index
The ratio of peak period to free-flow travel times, considering both reoccurring and incident delays ( e.g., traffic incidents).
Percent Travel Time In Congestion
Portion of peak-period vehicle or person travel time that occurs under congested conditions.
Congested Road Miles
Portion of roadway miles that are congested during peak periods.
Two times Free Flow
Evaluation of amount of peak travel time with is two times free flow travel time or more (generally used to indicate extreme congestion)
Travel Rate index
Used to indicate overall rate of progression by calculating the added time needed to make a trip under congested conditions summed across a network of roads
Benefit/Cost ( HERS)
Highway Economic Requirements System State - engineering/economic forecasting software used to identify possible highway problems and prioritise future investment. Uses traffic engineering data (speed, road length, volumes etc) as inputs to a model.
Buffer Time index
Weighted average for all sections of (95 th percentile travel rate mins/mile - average travel rate mins/mile)/(average travel rate mins/mile)%
Lost productivity Estimate (or Lost Efficiency)
Calculated by subtracting the peak period volume from the official capacity over a given time interval.
Estimate of how long congested "rush hour" conditions exist
Congested Lane Miles
The number of peak-period lane miles that have congested travel.
Annual Hours Of Delay
Hours of extra travel time due to congestion.
Oregon travel cost index
Contains a trade-off between the costs of land use and costs of delay, calibrated to favour compact land use. E.g. a 20 mins ride on a 2 mile road is favoured over a 20 mins ride on a 10 mile road.
Annual Delay Per Capita
Hours of extra travel time divided by area population.
Annual Delay Per Road User
Hours of extra travel time divided by the number of peak period road users.
Average Traffic Speed
Average speed of vehicle trips for an area and time ( e.g., peak periods).
Average Commute Travel Time
Average commute trip time.
Average Per Capita Travel Time
Average total time devoted to travel.
Notes to table
Source: author, from various sources
4.14 If the end use for an indicator is to provide information for transport users, then the public acceptability of a particular measure is an issue to be considered. Six different measures of congestion were presented to a group of 83 drivers of private and light commercial vehicles cars in DfT (2001) in order to asses their user value and acceptability as follows:
Table 4.3 - Alternative congestion measures to assess user acceptability
Basis for Measurement
Measures based on time lost per unit travelled for a typical journey and average vehicle
1) Secs/mile lost due to congestion
2) Mins/100 mile journey lost due to congestion
3) Hours/year lost due to congestion
Time spent in Jams (at standstill or speeds <mph)
4) % of time sent in jams
5) Mins spent in jams/hour of driving
Risk of serious delays
6) chances of serious delay
4.15 It may be worth noting that none of the above measures were well received by the sample of drivers questioned, but the time spent in jams was possibly most favoured. Measurement in terms of percentages or risk were perceived as most complex and least useful by the group. The notion that, in general, less complex indicators are favoured by both practitioners and travellers may be useful for future choice of indicators in the case for Scotland.
4.16 At a European level, a review of research has revealed a considerable programme of research concerned with congestion and road management from a system efficiency perspective, including SPECTRUM (1994), COSMOS (1996) and RECONNECT (2002). Much of this was undertaken within the early DRIVE programme of EU funded work, but related research has continued. Research has been concerned with the early prediction, detection and management of incidents in addition to optimizing the performance of the system as a whole. Formal definitions of congestion are difficult to identify, although the term is used widely within the research. One project with a formal definition is PRIME, which aimed to increase the effectiveness of incident detection and management on motorways and adjacent urban networks through the development of dynamic traffic management procedures. PRIME used the following as an indicator of congestion:
% change in Average Loop Occupancy Time per Vehicle ( ALOTPV) between periods with and without incidents
Wider impacts of congestion
4.17 In addition to the quantified indicators of congestion based around travel time or speed, research has shown that there are wider actual and perceived impacts of congestion, some of which are more difficult to quantify.
4.18 In a study aimed at improving the understanding of the extent to which accident risk increases in congestion for DfT (2003), despite an initial presumption that accident risk may increase in congested conditions, it was found that for urban and peri-urban sites, accident rates during periods of recurrent congestion are lower than those in uncongested conditions (less than half the accident rate 2). This was ascribed to the familiarity of regular road users with site conditions during periods of congestion and substantially lower speed of vehicles. Different results were found for motorway sites where the accident rate in congested conditions was nearly twice the rate in uncongested conditions; however the proportion of accidents that were fatal or serious was lower in congested conditions. For motorway sites the accident rate for Two Wheeled Motor Vehicles ( TWMV's) in congested conditions was found to be more than seven times the rate in uncongested conditions. For TWMV's, cyclists and pedestrians the proportion of fatal or serious accidents remained the same in urban and peri-urban congestion, probably reflecting the overall vulnerability to injury of these road user groups.
4.19 The perceived impacts of congestion were also discussed by DfT, 2001 as part of the qualitative findings from group discussions. These were reported on the basis of personal experience by car and light commercial vehicles from six areas in England involving travel of at least 2,500 miles per year and can be summarised as follows:
- Competitive or aggressive driving
- Driving found to be harder or more tiring
- Limited freedom of action or ability to travel where and when drivers wish
- Increased risk of accidents or mishaps
- Intensified pollution
- Increased fuel consumption
- Major source of driver stress - making many respondents feeling frustrated, angry, anxious, confused and/or exhausted.
4.20 These are consistent with other research findings, for example EU (2003). In moving forward towards a method of measuring the costs of congestion, both the quantified indicators and wider impacts of congestion have a role to play. Some aspects of the wider impacts are difficult to incorporate in costs and this is widely acknowledged - a typical example would be driver stress. The outline of methods to measure costs of congestion begins with a broad background in chapter 5, followed by more detailed descriptions the measurement of marginal costs in chapter 6 and total and excess costs in chapter 7.