3. What is Reliability and why is it important?
3.1 Journey Time Reliability versus Journey Time Variability
The impact of unreliability manifests itself in two parts: the level of unreliability, and the impact of each unit of unreliability. The former can be studied from past records and the interrogation of traffic data bases, while the latter will be estimated using literature search (see section 4) supplemented by an analysis of a suitably specified question in the survey of companies (see section 5).
This section will begin by teasing out some of the relevant concepts, and labelling them, firstly distinguishing between the Expected Journey Time and the Actual Journey Time. In statistical parlance, the former is just the arithmetic mean of the latter, and that definition will be adopted here. A journey undertaken in free-flow conditions can be taken to have the same Actual Journey Time on each occasion, and that will be equal to the Expected Journey Time. If planned road works forced a diversion over another road in free-flow conditions, the Expected Journey Time will increase. Avoiding such an increase will have a value to users, called the Value of a Travel Time Saving (VTTS), sometimes referred to more loosely as the Value of Time (VOT).
VTTS can be taken as made up of two parts. Firstly there are the costs of travelling for the extra time. For example, drivers might dislike spending time in slow moving traffic; bus passengers might dislike spending extra time on a bus; and slower freight movements will result in extra driver hours and vehicle operating costs. Secondly, there will be scheduling costs, for example: what utility might have resulted from having that time available to use in another way; and what disutility/costs might arise from having to set out earlier or arrive later? When considering changes in Expected Journey Times, it is implied that travellers can choose to minimise the impact on them by starting out earlier in cases where that is not prevented by some other fixed constraint (e.g. the time at which a play ends).
If journeys do not all take the same time, there will be Journey Time Variability (JTV), which can be measured (given sufficient data) as a standard deviation, for example. However, clarity is required over what dimension that variability is being calculated in. For example, journey times might vary over the hours within a day, but may be constant over days at a given time of day. Road users travelling at the same time each day would experience zero JTV, while those using the road at different times each day would experience a positive JTV. In both cases, however, it can be said that the travel times are reliable, as they are predictable from day to day, whilst being variable within the day.
The next consideration is whether the JTV could have been predicted before the journey began, in which case the start time could be chosen to minimise any adverse impact of unexpected delays. In the face of known JTV, travellers may be able to avoid the worst impacts of late arrivals by departing earlier, but may find themselves arriving too early at their destination on some occasions. In other cases, it will be impossible to start out earlier, both due to fixed constraints, and where the journey is already underway when the likelihood of delays becomes apparent. It is clear that disutility/cost is caused by JTV, and it is this that is usually valued as Journey Time Reliability (JTR).
3.2 Measuring Journey Time Reliability and its Value
The first (and sometimes only) step in measuring journey time (un)reliability is to measure JTV. Accordingly, the study obtained data on the variability of speeds, in 10 minute intervals, for the second Friday in each month of 2013, at selected trunk road sites across Scotland. The results from this work are reported in section 6. Regarding its value, VOR, a suitable question was included in both the Freight and Non-Freight surveys of companies. In the Freight survey companies were asked what costs would be incurred due to unexpected delays to a typical shipment, chosen by them. The Non-Freight survey asked what amount of delay once in every five journeys respondents would value equally to a specified fixed delay for all 5 days.
3.3 Strategies for Reducing Journey Time Variability
It follows from what has been said earlier, that not all JTV is worth removing, or even bad at all. However, there are cases where JTV reduction is likely to have significant value. Those are the cases where JTV leads to costly unreliability in arrival times. The first case is where the JTV within a day is so large, i.e. the 'peak' speeds were so much less than the free-flow speed, that there would be good grounds for imagining that the road would be subject to great instability of journey times in response to minor incidents or inclement weather conditions. This will be because the road is already operating close to its capacity at those times. In rural areas, considerable benefits might be obtainable by easing bottlenecks or generally providing more road capacity. In urban areas, benefits might be achieved by suitable traffic management measures – such as diverting some traffic, banned turns, reversible lanes, using variable message signs to alter speed limits, or queuing traffic in rear of locations unable to cope with the traffic offering.
The second case is where journey times at a given time of day and day of week are observed to vary from month to month. In this case the variability is probably due to either the weather, or varying seasonal demand for using that road. Rather than working with averages, road planning is improved if this variability is taken into account. Bad weather might slow down traffic seriously on a particular road, but that might increase demand on roads providing an alternative route, which should be catered for if possible. Routes that are busy just in the summer, with mainly recreational trips, might then suffer sufficient congestion that alleviation measures would pass Cost-Benefit tests, provided the appropriate data is used (i.e. not averages).
The data analysed in section 6 is for just one day per month, usually the second Friday, but this captured enough of the essence of travel time variability that the work will have wider lessons. Firstly, considering speeds in the peaks, the sample will have hit some holidays, which should not be ignored. Otherwise, it might be said that what has been captured is captured 'Day-To-Day' (DTD) variability (free of day of week effects), plus some seasonal effects. If a reasonable estimate of DTD variability can be isolated, it can be valued by inputting a suitable slack time allowance in all trips and a penalty for late arrival. Note that this speed variability will not be confined to the peaks. A much larger number of trips might therefore be affected, but probably to a much lower extent than in the peaks. Where the speed variability is great, actions such as those listed above for within-day variability may be appropriate. More likely is that amelioration would come in the form of improved strategies to deal with incidents, and to provide information to drivers.
In both cases, however, it is important to note that information for motorists should be managed. This will need to take into account what sources of traffic information are available to motorists, and actually being used by them at that time. For example, if the only information source was a single radio channel, and all drivers were tuned into that channel, then all drivers might follow any advice given simultaneously. If the incident in question is merely halving the capacity of a link, then it is not desirable to divert all the traffic onto an alternative route, but only a portion. Conversely, if an incident is considered too trivial to mention, then no drivers will divert. In response to a range of potential circumstances, there needs to be a range of possible responses to be chosen between - in the light of the current best guess of the proportions of drivers receiving various sources of information, and their perceived willingness to act on it. Rather than taking no action in the case of a minor incident, it might be worth advising just HGVs to divert to an alternative route. With increasing in-vehicle electronic information sources available, it should become much easier to smooth out the perturbations that do occur. By giving out carefully chosen information designed to optimise the system, drivers may come to trust the information provided, so that the crucial point will be to understand how each driver is likely to respond to that information. That will require a good understanding of traveller values of (uncertain) time in the vehicle and their value of late arrivals.
The study's work is naturally limited, but it throws light on the potential costs and benefits of taking this work further, should decision makers find that desirable.