Good day, and welcome to the Talking Operations web conference. Rich Taylor will be our moderate today. Please be advised that today's seminar is being recorded. I would like to go over a few logistical details. Today's seminar will last 90 minutes with 60 minutes allocated for the presenters and 30 minutes for questions and answers. If you think of a question you can type it into the smaller text box on the lower right side of your screen. Indicate to your question is directed towards, unless it is directed to all presenters and panelists. Please be sure that you are typing into the thin text box. Presenters will be unable to answer your questions during there presentations. If at any time you'd like to zoom in on a sled, you can click on the resume icon at the bottom left of your screen. It looks like a magnifying glass. Due to the size of the file, recorded files are available for viewing and listening only. At this time, our the to introduce Richard Taylor, our moderator. Rich Taylor, is a transportation specialist with the Federal Highway Administration's Office of Operations. He manages the operation's performance measures program. He is rarely assigned to the Office of Secretary to work on the congestion in the ship. Previously, he worked at ITS America. Thank you for joining us for today's webcast on travel time reliability. I have a few slides that I'm going to go through as an introduction. And then we have two experts who will give do presentations they will give you an idea of what they are and how they will be developed. And then we will finish of the Webinar with some questions from the audience. Please type your questions and I will be able to answer them. I'm going to start a quick presentation. Just to give you a little bit of history why there Federal Highway Administration is interested in travel time reliability. Basically, it came about because we're interested in trying to figure out how to measure mobility and congestion. In doing so couple would develop to the mobility monitoring program which uses archived data from traffic management centers around the country to actually develop some of the congestion and mobility measures. We used to provide an annual report of that data. We also, at the same time, came up with a monthly reporting program called the urban congestion reporting program, where we it originally used "webscraping" techniques. We also developed the same types of measures through that program. Developing these measures and looking at what kind of performance measures in terms of congestion and mobility and variability in travel times came about. That is how we got into the idea of travel time reliability. Today, what we are doing in the Federal Highway Administration is of the mobility monitoring program provides the state with agreements to smack the federally funded City Partner Program to provide traffic censors out in the field. That data is now provided in a monthly manner to the reporting program and we developed a monthly board. One of the measures that we currently report to this day planning time index. This is a travel time reliability measure. Shawn Turner, will give you an exact definition on that. I had a graphic here that is not high quality, but this is the overall executive summary of our monthly urban congestion report. This one is from May 2006. What it shows you is how some of the measures; we looked at travel time measures and the planning index. These measures that you are seeing on the screen report changes in these measures, month to month, quarter to quarter and year to year. I just want to show you an example of how these measures are actually being reported that. Wanted to mention that we released a day guidance document and per-share back and January. Both the reliability measures guidance document and brochure are available on the website you see on your screen. The idea is to get into greater detail and develop, you would use the measures for your particular agency. If you are interested in participating in these upcoming regional workshops, please contact me and we will work on getting you signed up for that. I also wanted to mention our annual report that was last released in September, and there will be another one released this fall. The report focuses on refining data that we collect there are programs. We also focus on the performance measures you are hearing about today. We also look at operational strategies. Here is an example of one of the congestion trends graphs that we have from our traffic congestion and reliability report cut just to give you some idea . Let me mention two other performance measures. They developed a list of 12 operations performance measures last summer. One of those measures happens to be travel time and reliability. As a follow-up, there is an NCHRP study getting under way this fall that will be piloting some of these measures. We are looking from the track for volunteers who would be willing to pilot those measures. That is something else we want to be aware of. We also have a reliability measures as well. That's just my introduction to the topic. If you want to contact me, my information is on the screen currently. Again, the weblinks for our travel time reliability information on our operations website is on the screen right now. With that, I would like to introduce our first speaker. Our first speaker is Shawn Turner. Shawn has conducted a wide variety of research since 1992. His emphasis has been in data collection, archiving and analysis. He currently leads several activities focusing on traffic data collection and performance reporting. His primary research sponsor and activities have been the Federal Highway Administration, the Texas Department of Transportation and the National Corporation of the Research. Thank you, Rich. It is a pleasure to be here, in my office, I guess, talking to everybody about travel time reliability measures. I am going to talk a little bit about the basics of travel time reliability measures and why our reliability measures important, what reliability measures we have developed and we think our useful, and how to calculate those reliability measures. We have chosen to focus on travel time simply because we have found that travel time is about the technical quantity as well as being something that a lot of nontechnical typical public audience and the media folks can understand and communicate to the general public. So, travel time reliability is not focused so much on typical delay as much as it is focused on unexpected delay. What the measures capture are really the variability, or it also captures some of those a really bad days. This is just a simple example. We look at a hypothetical commute. We look at it over three different months. In July, the travel times are very consistent between 13 in 14 minutes. The average is about 13 and a half. There is very little delay. The range is very small, so you could say that this Canute and July is a fairly reliable. Let's look at the second column, in November, again, the travel times range from 19-21 minutes. In this case, you got a fair bit of delay. It's taking you almost twice as long as it would during late or no traffic. In December, a typically there is a lot of other activity going on besides typical commutes. In this case, for this particular week in December, we've got a lot of variability. Travel times could be ranging from 13 - 24 minutes. The averages for November and December, they are probably about to 25 minutes. What I'm going to talk about our ways that we can quantify these differences in the reliability for different types of trips. Why do we think reliability is important to and why we should measure it. The first point is averages don't tell the full story. We have traditionally communicated traffic conditions by way of an average. In some cases, folks will only remember one number anyway. In some cases, an annual average or a monthly average may be inappropriate to. However, what people remember and what travelers experience can really be a lot different than the average that you may be reporting. If you look at the chart on the right-hand side over the course of the year, travel times vary. People remember those worst days. There is a problem of reconciliation in that too you're annual average may show something that really doesn't match up to the public perception of the situation. This really demonstrates a case where averages do not tell the full story. We need to have an additional measure that we can use to basically get at this day-to-day variability. The second reason that we think that reliability measures are important, and probably this is one of the most important reasons for folks in traffic operations and traffic management is that we believe that reliability -- Travel Time reliability measures in a lot of cases make better capture the large benefits of a traffic management and operation. That is to say traffic management activities may affect reliability in the long term more than they may affect averages. That is what I'm trying to demonstrate. On the left-hand side, you can see a lot of typical day to day variability. You institute a Freeway Service Patrol. Let me see if I can point to this. Or at least circle it. Your fluctuation has gotten less. You end up with a small improvement in average travel times. If we can look at reliability instead of averages and look at how you can decrease that day-to-day fluctuations, then certainly you would be able to better capture the benefits of some of those traffic management activities. The tax at the bottom of this slide is a case in point. Many of the folks may remember when the Minnesota Department of Transportation was required by the state legislature to essentially turn off their ramp meters...decline in travel time reliability. Clearly, in that case, and we think in a lot of other cases the travel time reliability measures better capture the reduction in day-to-day ability that is affected by a traffic management and operations activities. Let's talk about some of the reliability measures that we would recommend. Three of the measures are fairly related in terms of the fact that they use a 90th or a 95th percentile. We like to think of that 90 or 95th percentile travel time as being, for example, the worst commute of a particular month, with there being 20 work days in a month. Therefore, the 95th percentile would be that worse work day or that worst commute of the month. Again, trying to put some technical terms in terms of that the nontechnical this month. All lot of our measures use the 95th percentile. The selection of a percentile may, in some regard, depend on the type of trip or how critical a certain trip may be. The buffer index is a percentage of value that gets at how much extra time people have to build into their trips to make sure they are arriving early. For example, if I am driving across Houston, need to get to an important meeting, I am not going to budget to the average travel time to get to that important meeting. I am going to budget the average travel time plus some additional time cushion just in case their is a major incident and I get delayed by 50 or 20 minutes. What the buffer index does is it quantifies that extra buffer that would need to be added because of simply variable traffic conditions. Again, the buffer index is reported as a percentage and the planning time index is a value greater than one. It's essentially a multiplier that you would apply against your free flow travel time. Both of those are used in the 95th percentile. I was sure use some equations on the next several flights. Lastly, the fourth measure that we will talk about is simply the percentage of trips on time. This is similar to where exactly same measure that the airline industry reports in terms of on-time departures. In this case, we would report a percentage of trips that are on time. I have talked about the 90th or 95th percentile. Again, this is the travel time on some of the heaviest traffic days. It's reported in time and minutes. You would have to report separate troubled times for different origin destination pairs or possibly key routes. This percentile can be easily calculated in those databases or spreadsheets. Before I show you some specific equations for the Beaufort index. I wanted to show you this to better illustrate the 95th percentile and this concept of a buffer index. Now have found my printer. Hopefully you can see a number of trips along, in this case,eastbound in Seattle for the p.m. commutes. We have a whole range of travel times. Some of those trips along this particular corridor had a reasonably good travel time. If we take and calculate the average travel time, the average time is about 16 minutes. Let's move out a little bit further. This is out at this end of our curve. These are the most painful commutes where the travel times are two or three times what you might expect. For example, the 95th percentile is about 23 minutes. What this illustrates is the gap between the average travel time in the 95% drop. If by budget the trip, half of the time I'm going to be early. Half of the time, I'm going to be late.That difference is essentially what we refer to us a offer index; the buffer index is reported as minus the average travel time and /D average travel time. I comes out as a percentage. The nice thing about reporting it as a percentage value is you now have the ability to compare the buffer index across several different as night because you have of normalized your travel time. You have combined the buffer into a regional average simply by taking a weighted average of the individual. I am also showing on this slide that if you would like you can report the buffer time or that extra amount of time cushion. You can report that as well in minutes by multiplying the buffer index by the average travel time. Another concept that is a slightly different than the total time that a traveler should allowed to ensure on-time arrival. The buffer index represents the extra time that should be added to the actor, where is the planning time represents the total time. For example, the total time might be used more for travel information as opposed to an incremental or additional time. The planning time is -- / the ideal or free flow travel time. That results in an index value that is greater than or equal to and as simple to so it can be abbreviated. This slight, hopefully -- shows the relationship between four average travel time. This is a typical time of day profile. In this case, what we are showing in the yellow line is the travel time index. It is simply the peak or the current travel time. This. Represents average conditions. Averages don't tell the whole story in does not capture the day-to-day variability. What we can do then is essentially showed the planning time index for the 95%. This red line represents the 95th percentile. It essentially shows you near worst case travel times. The planning time index represents that upper bound on-time reliability. The fourth thing is the percentage of trips that are on time. This measure, probably the most critical decision is what is this value going to be, essentially what do you want to consider a trip to be on time, or how much to lay do you consider to be acceptable in a travel time to still be within a reasonable bound? This is the same concept as the on-time reliability that is reported in air travel. I have mentioned this earlier, but I wanted to reinforce that when we are talking about travel time reliability, we want to put it and terms that nontechnical people can understand. Have shown here on this slide the standard deviation and coefficient. Those are statistical measures of variability. It's difficult for members of the general public, even commuters, to understand what a standard deviation is or to relate a standard deviation to their typical commute. These are going to be difficult to to communicate as well as for folks to understand. Richard Taylor, and his introduction had indicated a report that was produced and released in January, of this year. It is called "Travel Time Reliability, Making it There On Time, All the Time." You can start at the operations website and look in their publications area, or you could use your favorite search engine. We essentially have a brief brochure as well as a larger technical document that essentially gives folks some guidance on how to develop travel time reliability measures. Certainly, I have Rich's name and my name listed up there. If you have any questions, feel free to contact us. With that, I think we are saving questions. I will turn my presentation over to Rich. Thanks, Shawn. Our next speaker is Rich Margiotta. As soon as I get his bio up, I will tell you all about him. Rich is a principal with Cambridge Automatics. He has held positions at SAIC, the Transportation Research Board and Dismay. He received his Ph.D. in civil engineering from the University of Tennessee and a bachelor's degree in biology. His recent work is concentrated on performance measures for congestion and operations. Welcome, Rich. Thank you, Rich. It is a pleasure to be here. I'm going to pick up where Shawn left off, give some examples, both hypothetical examples of how reliability could be used as well as some real world examples where reliability has been applied. To be frank, their has not been a lot of applications developed in the trenches with reliability, since it is such a new concept. We expect the number of applications to grow. Shawn has really already talked about this, the basic definition of reliability. I am not going to spend any time on this. I will mention that from an economic perspective, the amount of extra time that has to be planned for to account for not being late for the buffer, as Shawn put it, has a cost associated with it. Some economics studies have been at a rate much higher than what the average would be. There are additional costs associated with the buffer. There is some economic evidence that this is placed at a premium by travelers. One of the main points I want to make three my presentation is that reliability basically says you have inconsistency or variability in troubled times due to something. It doesn't tell you what is causing it. What are these causes of congestion? I've listed seven potential sources. These are listed here. Typically, related events that cause trouble times to be abnormal. Noticed number seven, the problem is that base capacity is largely determined by what the impact of a particular event is going to be. And if you have the floor lay into a delay block one, you have less of an impact because you have excess capacity capable of observing the effects of that event. It is one thing to just list the seven sources and say these are the things that caused trouble. Wouldn't it be nice if we could just break those out and some them all out. This first slide shares are traditional highway capacity view. We have a physical capacity that relates to a static demand volume and we get an estimate of what the delays are going to be; variability in traffic demand as well as special events. Demand itself is not static. In addition, when we start throwing additional events, it gets a lot more complicated. The amount because additional incidents. All of these factors are interrelated and produce what we tend to think of as total congestion. These factors are highly interrelated. That all contribute to congestion. Pricking these out into individual estimates is not as easy as it might sound at first glance. Let's look at some additional data here on reliability. This is the history of travel time in Seattle, taken from detectors and the roadway. You can see on some days, traffic is a lot lighter than others. On other days, we have incidents and some days we have incidents with bad weather as well. You can see what these different events are doing in terms of causing congestion. Some examples of how things have changed over history, here is an example of a very highly congested corridor in Atlanta. It is only four miles in length. You can see that the average congestion has grown. Breaking down liability into its sources is really one of the primary aspect in my mind. It can lead to measures and actions. This particular graphic was done by consensus, taking a look at some of these studies that had been done over the past ten or 15 years and coming up with a very rule of thumb sort of estimates. In the last three or four years, their have been some empirical studies focusing on individual freeway corridors, where data from the roadway and the events have been merged to try to get a more precise estimate of what the various contributions to each one of these sources really is. You can see that these studies are all in a fairly broad range of estimates in terms of the percentage of contributions. It really depends on the nature of the quarter. It is that bad weather? Do you have a lot of incidents? To you have places where incidents can be pulled off? It really gets down to the individual level. Also, one of the things that I would like to point out is when you start getting a high level of base congestion, bottleneck related congestion, that tends to dominate congestion in an individual corridor. All of these particular course, with the exception of Albany, were bottleneck intensive corridors. They were all fairly large, but when you get down to the highly congested individual corridors, the bottleneck the to late tens to override these incidents in terms of... It may be a lower percentage than the national average, but it is a huge number in terms of delays. What I the effects of treating reliability? Here is my favorite corridor in Atlanta, because it is though heavily congested. This is actual travel times. We said let's take all of the extreme travel times and just across the border reduce them by 25%. That produces the shift to the orange line. Shawn noted this in some of his previous examples where you not only get a reduction and reliability, but the distribution is a lot tighter. If you are treating incident to lay or special event related delay, you are going to have an effect on total delight. You get this double benefit by treating the events that cause unreliable travel. I noticed a question that popped up and what is useful to individual travel. What is it going to be like today? Here I just plodded some travel times. The 95th percentile is shown here. It's getting up into the range of the worst 5% days of the year. Placing what is happening right now in the system in historical context can give travelers some additional information on how to react. It gives them basically a baseline or a reference .5 plot think current conditions vs historical averages and reliability estimates. One of the major uses of reliability measures is an ongoing performance measurement program. Again, this is the notion that overall reliability tells you that you have the variability somewhere, but if you drill down into the individual sources, you can build this cascading from work for performance measures that eventually leads to action. Let's walk through this from top to bottom for example. My overall reliability measure is large. Compared to other corridors and cities. That happens to be large. By start drilling down, I can take all the components that make up incidents, all the times related to the incident. Response time, by comparing this, I can determine bracken best focus my resources. Using reliability in this context helps you -- provide you a basis for drilling down to the level at which actions can actually be undertaken. One of the concepts that we are promoting this this idea of a travel time reliability profile. This is meant to be a one-page snapshot of the reliability for individual corridors. We've got our old friend here, the distribution of travel times. What we have done is superimposed the reliability and congestion metrics on top of it. In addition, we are also trying to understand what the nature of events are on this quarter. The bottom chart here which could easily be a pie chart provides estimates of the percentage of contribution to total congestion for the individual sources. In the box shows some of the individual characteristics in this particular quarter. Just another example of another highway. This was a three hour peak in Seattle. Same sort of affair. I will talk briefly about some of the other applications and studies that have been undertaken in the use of a reliability. I have listed several here. The Southern California Association of Governments did a planning level steady where they wanted to estimate the economic impact of on reliable travel. They came up with some estimates of what the reliability was and what the economic cost of that is. Basically, by taking some fundamental relationships. As we compile data from several different cities, as congestion level rises, that relates back to the basic model as base capacity starts to disappear. Here is an example. They are taking readings from the satellites positions of trucks along several corridors throughout the country to get an estimate. Finally, this is a companion NCHRP too. It was developed several years ago where the research gaps were on the planning and operations this month. The shark to program says perhaps we need to have a more strategic and concentrated approach to develop and research and planning and operations. One of the four areas for the sharp two program is called reliability, along with the other three areas that I've got listed here. This program as still being developed, but it is real and it was funded. Currently are listed a total of 16 projects that will be undertaken throughout a four year period. That number may increase slightly or decrease slightly, but it will roughly fall into these individual categories. These will be multiple year, very we'll funded projects to get at some of the core issues that have been nagging us for the past 20 or 30 years in terms of being able to model reliability as well as the impacts of various operational strategies are. That program will probably have its first contract issued sometime this fall. Again, it is a four year program. Reliability was considered to be important enough to be included as one of the four focus areas. Finally, you've probably seen some of these references before it will be a manual or a cookbook to displace some of these reliability measures in such greater detail than we could get into today. Washington state DOT has been a leader in travel time and congestion metrics and reliability as well. I encourage you to go to their site to see some of the graphics that they have developed. They are really cutting edge, state of the art graphics that they have developed there. I believe that is it. Thank you very much, Rich. I want to thank both John and Rich for their informative and very detailed presentations. We have a good list of questions to get to. I wanted to make a couple of comments to begin with. First of all, back in Shawn's presentation when he was talking about the Minnesota Ramp Project, he said there was a decline in the travel time reliability, it makes it sound bad, but that's actually a good thing. I wanted to mention that in case anybody got confused by that metric. It just meant that the number went down. That was a good thing to happen. I think what it did is it demonstrated that ramp meters and was playing a significant role and when you turned it off, things got really bad pretty quickly. I'm going to go through our questions in the order we got them. Feel free to jump in if you have the answer. There was a question about the urban stuff that I mentioned in the introductory slide. I wanted to mention how books became a part of that. Through the Mobility Monitoring Program, and Shawn and Bridget both work on that program. They go out and talk to states and local urban areas around the country to see what the status of their archive data is available at the end of the month. You have to provide that data to the monitoring program. We have a limited amount of resources to use. Right now, we are limited to 20 cities, although we are still looking to add additional cities as they become available and funding permits. Basically, if you are interested in getting your city to be part of the reporting program, you need to talk with Shawn or Rich and talked about your capabilities with them. Right, what it would require is something pretty close to a daily data feed that has all of your traffic detector data. One of the other things to have done is much and programs. Ultimately, the best thing that could come out of it is that a state or region develops their own performance monitoring program that they can use for their own decision making. Any prototypes you would like to see done, and a customized data analysis, within reason, it's within our charter to do so. Our first question, let me go back up here and read. Regarding consolidating they are the understandable, easy-to-use and make sense? We describe it right now for more the internal program on their website where you enter where you are coming from and where you are going and a key issue the travel time. What it is saying is that typically this trip made take 12 minutes but it could take as much as 24 minutes. That is the best well have seen these measures being used for the actual use of the traveling public. I don't know if either of you want to add to that at all. The question was our their efforts to validate with the traveling public that the reliability measures are understood and make sense. The answer to that is no, their has been no formal validation effort where we have done focus groups or anything like that.