Travel Time Reliability Analysis

Travel Time Reliability Analysis CHAPTER TWO Literature Review 2.1 Introduction Lyman (2007) states that travel time reliability is vital measure of congestion and can serve as benchmark for prioritizing improvements into a city transportation system. This research start with a literature review of travel time reliability and its worth as a congestion measure. Travel time reliability can be denoted as the probability of successfully completing a trip within specified time interval (Iida, 1999). Therefore, the increase of travel time will lead to the unreliability and variability of travel time (Recker et al., 2005). The better understanding of travel time reliability and variability might assist transport planner to select proper transport policy in conjunction with reduction congestion problems as well as lessening the impact of different type of incidents (Recker et al., 2005). It can be said that, the more reliable the transportation system, the more stable is the performance. In addition, lower travel time fluctuation also contributes to less fuel consumption as well as less emissions due to a reduced amount of acceleration and deceleration by vehicles (Vlieger et al., 2000). Moreover, from a transport users point of view, more reliable travel times mean more predictable journey times and improved activity schedules. In accordance with just in time services, reliable travel time will significantly increase the freight industrys performances to deliver goods (Recker et al., 2005). As travel time reliability considers the distribution of travel time probability and its variation at road network, the higher travel time variance the lower travel time reliability (Nicholson et al., 2003). It can be also said that under ideal conditions travel time reliability would have a variance equal to zero. Indeed, the increase of its variance will therefore significantly reduce its reliability. However, the relationship between travel time variance and its reliability is not linear, so that, it cannot be generally accepted that a double of travel time variance will lead to a half of its reliability. To conclude, the greater travel time fluctuations will have significant impacts on transport network reliability. According to different purposes of travel time reliability study, there are several travel time reliability surveys. By comparing different aspect of the travel time study and by considering the complexity of data collection as well the data analysis, Lomax et al. (2003) has reviewed the suitable assessment of travel time reliability. Based on the scope and the limitation of each method this work suggested the different study in terms of measurement travel time variability and travel time reliability. The analysis of the archive traffic data is not proper in measuring the travel time reliability due to the lack of data constant and the lack of other attribute related with the traffic condition. However, the data is easy to obtain. In addition, the micro simulation techniques have been used extensively, however according to Lin et al (2005) there are some deficiencies in travel time micro simulation modeling in terms of the high need for data calibration. In order to gain real life traffic conditions, some travel time reliability research used the probe vehicle methods. Since this method requires ext ensive labour and only covers some of the study area or some of the road segments, it cannot be applied in terms of assessing the travel time reliability on large road networks. Indeed, Lomax et al also recommended some reliability measurements by examining the reliability and variability percentage (e.g., 5%, 10% and 15%). Those approaches take into account the effect of irregular conditions in the forms of the amount of extra time that must be allowed for travelers. The first measurement is the percent variation which expresses the relationship between the amount of variation and the average travel time in a percentage measure. The second is the misery index that calculates the amount of time exceeded the average slowest time by subtracting the average travel time with the upper 10%, 15% and 20% of average travel rates and the last is travel time buffer which add the extra travel time of 95% trips in order to arrive on time. In addition, since reliable travel time is the key indicator of users route choice there are many recent research works which investigated the travelers behaviour under unreliable travel time. According to travelers behavior in route choice survey, the greater the variance of travel time of selected links the less attractive it is (Tannabe et al., 2007). Additionally, Bogers and Lint (2007) investigated traveler behavior on three different road types in The Netherlands under uncertainty conditions, as well as the impact of providing traveller information on route choice. They conclude that providing traveler information has significant impact on effecting travelers decision, in addition, based on travelers experience they will choose the route with minimal travel time variance. It means that the routes that have high travel time reliability are not attractive for users. Indeed, according to Lomax et als review that the best alternative to measure the travel time variability and route choicer behaviour under uncertainty condition is by using probe vehicles. Though this method was highly labourious and expensive, it is more realistic (Lomax et al., 2003). Then Tannabe et al (2007) undertook an integrated GPS and web diary in Nara, Japan. This study found that travelers might change their route to reduce the uncertainty in travel time. In addition, there was a positive correlation between coefficients of variation (CV) of the commuting routes. It is found that the appropriate functional hierarchy of road may be disturbed by the uncertainty of travel time. These findings suggest that a reliability index of travel time is very useful and important for evaluating both actual level of service (LOS) and functional hierarchy of road network. Recent travel time reliability research investigated the relationship between the traveler behavior and their response to the provision of travel information system while they experience high travel time variability. Asakura (1999) concluded that the Stochastic User Equilibrium model can generate the user route choice behavior based on the different levels of information provision. This study analyzed two different groups, the first group being the well informed users and the second the uninformed users. He concluded that providing better information can improve the transportation network reliability. In order to find out the different perspectives of travel time reliability for different persons with different purposes, Lo et al (2006) studied the notion of the travel time budget, in which each traveler seeks to minimize their own individual travel time budget (the amount of time that the individual is prepared to devote to travelling), which means the total travel time of the individual should not exceed their allocation of time to travel. To evaluate the link between the presence of ramps on motorways and travel time reliability, recent reliability network research has been undertaken in The Netherlands. Th is study analyzed whether the geometry of road network also affected the travel time reliability (Tu et al., 2007) by investigating the presence of ramps on six major. This study concluded that the presence of ramps in the road network has reduced the travel time reliability. Since road network reliability considers the probability of transportation system failures in how to meet performance parameters such as reasonable travel time and travel cost, level of service and the probability of connectivity of the transport network and lack of measuring the consequences of link failure to the community, the concept of road network vulnerability might be an alternative way to fill some of road network reliability deficiency, particularly in assessing the adverse socio-economic impact to community (Taylor et al. 2006). ROAD NETWORK VULNERABILITY Due to the potential socio-economic cost of degraded transport network to community, the concept of road vulnerability has been developed by researchers under transport network reliability umbrella. The definition of vulnerability has not yet been generally agreed. Several authors notion of the vulnerability focused on the negative events that significantly reduced the road network performance. Berdica (2002) defined the vulnerability as a susceptibility to incident that can result in a considerable in road network serviceability. The link /route/road serviceability described the possibility to use that link/route/road during a given period of time. Furthermore, since accessibility depend on the quality of the function of the transportation system, this concept relate to the adverse of the vulnerability in terms of reducing accessibility that occurs because of the different reasons. As the idea of network vulnerability relates to the consequences of link failure and the potential for adverse socio- economic impacts on the community (Taylor et al., 2006, Jenelius, 2007a), thus vulnerability can be defined in the following terms: 1. A node is vulnerable if loss (or substantial degradation) of a small number of links significantly diminishes the accessibility of the node, as measured by standard index of accessibility. 2. A network link is critical if loss (or substantial degradation) of the links significantly diminishes the accessibility of the network or of particular nodes, as measured by standard index of accessibility. Therefore, it can be concluded that road vulnerability assesses the weakness of road network to incidents as well as adverse impacts of the degraded road network serviceability on the community. In relation with the road network vulnerability definition which focuses on two different aspects; selecting critical road network elements and consequences of measurements, Jenelius (2007a) has identified that road network vulnerability assessment can be distinguished into two stages. The first stage is to select a critical link by identifying the road network likelihood and by quick scanning of wide road transport and the second one is measuring the consequences of link disruption to community. Based on previous works, different approach has been applied in order to scan wide road network. Jenelius et al ( et al., 2006) selected particular major arterial road which connect the district at the Northern Sweden to be the worst case scenario and selected road links randomly as the average case scenarios. Scott et al (2006) has also introduced topology index and the relation between capacity and volume then select the critical link. Indeed, Jenelius (2007a) has suggested that conducting comprehensive assessment of road network will be helpful for identifying roads that are probably affected by the traffic accident, flood and landslides. Berdica et al (2003) undertake a comprehensive study in order to test 3 types of software to mode l road network interruptions. This study simulated the short duration of incidents on University of Canterbury networks by using SATURN, TRACKS and Paramics. They modelled a total block of one link on the small network then run the model at the macroscopic level by using TRACKS, at mesoscopic level by using SATURN and at the microscopic level by using Paramics. Based on the simulation, the different packages gave different result in terms of their responsiveness to model the short incidents, for instance, Paramics might be considered as a suitable software package for short duration incidents because it is more responsive than other softwares. SATURN which is more detail in its formulation than TRACKS has less responsiveness than TRACKS. Given the lack of generally recognized measurement of road vulnerability, it has been common practice to consider measures such as the increase of the generalized travel cost, the changes of the accessibility index or the link volume/capacity ratio when one or more links were closed or degraded as road vulnerability measurement. Taylor et al (2006) studied the network vulnerability at the level of Australian national road network and the socio economic impact of degradable links in order to identify critical links within the road network, by using three different accessibility approaches. The study introduced the three indices for vulnerability. The first method was the measurement of the change of the generalized travel cost between the full network and the degraded one. This method has concluded that by degrading one particular link the generalized travel cost will increase, and then the links which gave the highest travel cost was determined as the most important link. The second method used the changes of the Hansen integral accessibility index (Hansen, 1959) in order to seek the critical links. It was assumed that the larger the changes were after cutting one link, the more critical that link was on the basis of the adver se socio-economic impacts on the community. The last approach considered the changes of the Accessibility/Remoteness index of Australia (DHAC, 2001). This method was similar to the second method which sought the critical link depending on the difference between the ARIA indices in the full network and the ARIA indices in degraded network. Moreover, Taylor et al (2006) also studied the application of the third approach at the regional level in the state of Western Australia. This study concluded that removing a link gave different impacts for the cities, for example, by cutting one link, the impacts on the several cities were only local, in contrast, other cities where they were available similarly alternative road performance did not give significant changes of the ARIA indices. Due to the importance of a particular link within the wide road network, Jenelius et al (2006) introduced a similar approach to Taylor et al (2006). They studied the link importance and the site exposure by measuring the increase in generalized travel cost in the road network of the Northern Sweden where the road networks were sparse and the traffic volumes were low. By assuming the incident was a single link being completely disrupted or closed so the generalized cost increases, then the most critical link of the operation of the whole system and the most vulnerable cities because of the link disruptions were determined. The study concluded that the effect of closing a link was quite local and the worst effect was in the region where the road network was sparser with fewer good alternative roads. This research suggests that the road network vulnerability assessment can be applied in road network planning and maintenance, to provide guidance to the road administration for road prioritization and maintenance. In addition, Taylor (2007) studied the road network vulnerability in South Australia road network which included all the freeways, highways and major main roads. This research used a large complex road network and evaluated the ARIA indices changes for about 161 locality centers with populations exceeding 200 people. This study found the top ten critical links in the South Australia regional road network. Moreover, in relation with vulnerability approach in D Este and Taylor (2003), Chen et al (2007) tries to assess the vulnerability of degradable networks by using the network based accessibility and by combining with a travel demand model. Their study concluded that themodel can consider both demand and supply changes under abnormal conditions. Thus the vulnerability network assessment can be measured by considering the duration of the disruption (increase the travel time) and modeling the user equilibrium both the cases when there are alternative roads or the case when there are not alternative roads (Jenelius, 2007b). Indeed, Scott (2005) introduced the measurement of the Network Robustness Index by considering the ratio between the link capacity and link volume and assigning topology index for each link then test whether the particular links can cope with the changes of the traffic demand when one or more links were closed or degraded (Scott et al., 2005). Jenelius (2007b) introduced the new method in order to incorporate dynamic road condition and information by assessing the increase travel time using the extended of the user equilibrium model. This study assumed that there was no congestion and there was at least one alternative route between the origin and destination. Further, this study also assumed that the road users have perfect road information about the length of road closure so that they can decide whether they need either to take a detour or to go back to their origin and wait until the road reopened. This method calculated the additional travel time which is calculated since the road users were informed about the road closure, the waiting time until the road reopened. The difference between the normal travel time and t he additional travel time due to road closure was assigned as the increase travel time. However, this study did not take into consideration the change of the travel flow at the alternative routes. This assumed that the mix of the current and diverted traffic can flow at the free flow. In order to assess the increase of the flow when the diverted traffic mix at the current traffic which already meet the capacity or are already congested, the study which conducted by Lam et al (2007) can be considered. This method introduced the path preference index which is the sum of the path travel time reliability index and the path travel time index. To examine road network vulnerability in an urban area, Berdica et al. (2007) studied the vulnerability of the Stockholm road network by examining 12 scenarios involving partial and total closure of selected links, including bridge failure. Also, it assessed the road network degradation in three different times of day, morning peak hour, middle of d ay and afternoon peak hour. This study concluded that by closing one link or all links as well as bridge failure would increase the total travel time and total trip length (on the assumption that travelers chose their minimum time route based on user equilibrium method). The model of different scenarios at different times gave different results but the most vulnerable links were the Essinge route and the failure of Western bridge scenario. To conclude this study calculated the increase of total travel time a day and then multiply that by 250 days to obtain the total increase travel time for yearly basis. Though the highest total travel time increase in only 8% per day, however if it is calculated by 35 SEK (travel cost per hour) it gave significant impact of total travel cost increase. However, it did not take into account the duration of the closure and left some discussion of link disruption impacted such as the effect of noise and pollution during the road closure. Moreover, Knoo p and Hoogendoon (2007) assess the spillback simulation in dynamic route choice in order to examine the spillback effects then evaluated the road network robustness and the vulnerability of links. This study concluded that it is necessary to assess the spillback effect in order to identify the most vulnerable link within the wide road network. Tampere (2007) investigated the vulnerability of highway sections in Brussels and Ghent. This work was quite challenging, it tried to consider the different aspect of the road network vulnerability criteria related to the amount of vehicle hours lost due to major incidents. This work compromised of two steps; the first one is the quick scanning of the most vulnerable link from the long list into short list by considering the several aspects and then by obtaining the short list links then the vulnerability measure was conducted. Since this method used the dynamic traffic assignment, there are some drawbacks during the model run such as the lack of traffic distribution after the occurrence of the incident which resulted an illogical of travelers route choice. In general this method has successfully measured the vulnerability by not only considering the traffic condition but also taking into account the different road networks. Though this method has not considered traffic assignment criteria, it is still considered as a refinement over similar studies Measures of Congestion used in Transportation Planning Measures of congestion are intended to evaluate the performance of the transportation network and to diagnose problem areas. They provide information on how well the system has met certain stated goals and targets, and can also help to explain variations in user experiences of the system. There are four general categories of congestion measures. The first category contains measures that explain the duration of congestion experienced by users in some way; these include delay, risk of delay, average speed, and travel time. The next category includes measures that analyze how well the system is functioning at a given location. This category primarily consists of the volume to capacity (V/C) ratio, which is usually expressed as a level-of-service (LOS) category. LOS is a performance rating that is often used as a technical way to express how well a facility is functioning. For example, a facility functioning poorly is likely to be rated as LOS F, but could just as easily be described as poor. The third category is that of spatial measures, including queue length, queue density, and vehicle miles traveled. It is important to note that some of the duration and spatial measures are actually measured as point measures. The final category of measures is the other category, consisting primarily of travel time reliability and the number of times a vehicle stops because of congested conditions. Easily the most common measure of traffic congestion is the volume-to-capacity ratio. The V/C ratio measures the number of vehicles using a facility against the number of vehicles that the facility was designed to accommodate. This ratio is an important measure for planners to use, and represents an easily understandable measure of whether or not a roadway is congested. However, it can lead to some philosophical problems, such as whether transportation systems should be built to handle the highest demand or the average demand, and what level of service is acceptable. In addition, it is difficult to accurately measure the capacity of a roadway. The volume-to-capacity ratio is an important tool for comparing a roadways performance to other roadways and over time, but does not necessarily reflect the overall user experience and values in the system. Despite the prevailing usage of the volume-to-capacity ratio, and perhaps because of its inherent philosophical difficulties, the (FHWA) ha s strongly encouraged agencies to consider travel time experienced by users as the primary source for congestion measurement. They also state that currently used measures of congestion are inadequate for determining the true impact of the congestion that clogs up the transportation system from a users perspective, and that they are not able to adequately measure the impacts of congestion mitigation strategies. What is travel time reliability? As mentioned in section 1.1.1, the OECD (2010) provides a general definition for Travel Time Reliability: The ability of the transport system to provide the expected level of service quality, upon which users have organized their activities. The key of this definition is that a route is reliable if the expectations of the user are in accordance with the experienced travel time. But this does not directly lead to a TTR measure. Nonetheless, this definition shows that user expectations should be taken into account when selecting a proper TTR measure. Congestion is common in many cities and few people will dispute this fact. Drivers become used to this congestion, always expecting and plan for some delay, especially in peak driving times. Most drivers budget for extra time to accommodate traffic delays or adjust their schedules. Traffic delays are mostly much worse than expected when it happens. All travelers are less tolerant of unpredicted delays, the effect is that it makes then to be late for work or vital meetings, miss appointment, or suffer additional childcare fees. Shippers and freight forwarders who experience unpredicted delay may lose money and interrupt just-in-time delivery and manufacturing processes. Traffic congestion used to be communicated only in terms of simple average in time past. Nevertheless most travelers experience and remember a different thing than the simple average as they commute within a year. Travelers travel time differ from day to day, and remember the few bad days they suffered through unexpect ed delays. Commuter build time cushion or buffer in planning their trip to account for the variability. The buffer helps them to arrive early on some days, though not a bad thing, but the additional time is carved out of their day time which could have been used to pursuit other activities than to commute. Travel time reliability time frames Travel Time Reliability can be categorized by its time frame. Bates et al. (2001) discusses three levels of variability: inter-day, inter-period and inter-vehicle. Martchouk et al. (2009) explains these as follows: Inter-day: Variations in the travel time pattern between days. Some days of the week might have substantially different traffic volumes than others. For example, a Sunday will generally have less traffic than a Monday. Same weekdays should have about the same travel time pattern, but there can still be variations. Also, events such as road works or inclement weather cause inter-day variations. Inter-period: Variations in travel times during a day. Many road sections have a morning and evening peak, during which travel times are larger. These variations are caused by variations in traffic volume. Inter-vehicle: Relatively small differences in travel times between vehicles in a traffic stream. These are caused by interactions between vehicles and variations in driver behavior, including lane changes and speed differences. Although Martchouk et al. (2009) shows that individual travel times on a motorway section can vary strongly in similar conditions, due to driver behavior, this study focuses on inter-day variations. It is assumed that inter-vehicle variations have no significant influence on Travel Time Reliability. In urban areas, the speed difference between vehicles will generally be smaller than on highways. The reasons for this are: the average speed on highways is higher, there is more overtaking, trucks cannot drive at the maximum allowed speed, and routes are longer. Inter-period variations are also not considered, because it is presumed that road users know that travel times within a day vary according to a more or less fixed pattern. It is the deviations from this daily pattern which are interesting in the light of TTR, since these cannot be predicted by road users. Therefore, the focus of this investigation is on inter-day variation. Why travel time reliability is important? Travel time reliability is vital to every user within the transportation system, whether they are freight shippers, transit riders, vehicle drivers and even air travelers. Reliability allows business travelers and personal to make better use of the own time. Because reliability is so significant for transportation planners, transportation system users, and decision makers should consider travel time reliability as a key measure of performance Traffic management and operation activities is better quantified and beneficial to traffic professionals by the use of travel time reliability than simple average. For instants take into consideration a typical before and after study that attempts to quantify the benefits of an accident management or ramp material program. The development in average time may seem to be modest. However reliability measure will show a much greater development because they show the effect of improving the worst few days of unexpected delay. The Beginning of Travel Time Reliability as a Performance Measure Hellinga (2011) states that in the past, analysis of transportation networks focused primarily on the estimation and evaluation of average conditions for a given time period. These average conditions might be expressed in terms of average traffic stream speed; average travel time between a given origin and destination pair; or some average generalized cost to travel from an origin to a destination. This generalized cost typically includes terms reflecting time as well as monetary costs. These terms are summed by multiplying the time based measures by a value of time coefficient. A common characteristic of all of these approaches is that they reflect average or expected conditions and do not reflect the impact of the variability of these conditions. One reason for this is that models become much more complicated when this variability would be included. Also, a vast amount of data from a long period of time is needed. Unfortunately, collecting data is often costly and time-consuming. H ellinga (2011) also observes that more recently, there has been an increasing interest in the reliability of transportation networks. It is hypothesized that reliability has value to transportation network users and may also impact user behavior. Influence on traveler behavior may include: destination choice, route choice, time of departure choice, and mode choice. It is useful for road managers and planners to have knowledge about the relations between TTR and road user behavior, because this can be used to predict or even deliberately influence this behavior by applying traffic management measures. Consequently, there has been an effort to better understand the issues surrounding reliability, and to answer a number of important questions such as: 1. How is transportation network reliability defined? 2. How can/should network reliability be measured in the field? 3. What factors influence reliability and how? 4. What instruments are available to network managers, policy makers, and network users that impact reliability and what are the characteristics of these causal relationships? 5. What is the value of reliability to various transportation network users (e.g. travelers, freight carriers, etc.) and how is this value affected by trip purpose? 6. How do transportation network users respond to reliability in terms of their travel behavior? (E.g. departure time choice, mode choice, route choice etc.) 7. How can reliability (and its effects) be represented within micro and macro level models? (Microscopic models focus on individual vehicles, while macroscopic models pertain to the properties of the traffic flow as a whole.) 8. How important is it to consider the impact of reliability in transportation project benefit/cost evaluations? 9. Does the consideration of the impact of reliability within the project evaluation process alter the order of preference of projects within the list of candidate projects? Hellinga (2011) states that the above list of questions, which is likely not exhaustive, indicates that there currently exists a very large knowledge gap with respect to reliability. Various research efforts around the world are beginning to fill in these gaps, but the body of knowledge is still relatively sparse and there is not yet even general agreement on terminology. Note that the first, second, and (partially) fifth question are part of this investigation What measures are used to quantify travel time reliability? The four recommended measures includes 90th or 95th percentile travel time, buffer index, planning time index, and frequency that congestion exceeds some expected threshold. These measurements are emerging practices, some of