Title: Estimating Benefits of ITS Technology in Incident Management: The Case of Northern Virginia Author(s): Gerard C. Maas, Jr. Research Fellow Transportation Policy Program The Institute of Public Policy George Mason University Fairfax, VA 22030 Phone: 703-993-1320 Fax: 703-993-2284 Email: gmaas@gmu.edu Transportation Research Board 77 th Annual Meeting January 11-15, 1998 Washington, D.C.
Maas 1 INTRODUCTION Metropolitan areas across the nation are faced with increasing congestion. Non-recurrent congestion from traffic accidents and other incidents is a major component of the overall congestion problem, and is being addressed through incident management programs. In view of pending deployment of new technologies, incident management must be linked to ITS. Building a mandate and obtaining sufficient funding for ITS-enhanced incident management is often hampered by the inability to satisfactorily quantify the expected. The focus of this paper is therefore on the issue of estimating benefits of ITS technology application in incident management. In the first section, the background issues of congestion, ITS and Incident Management response and benefit estimation are discussed. These issues are placed in the specific regional context of Northern Virginia in the second section. A brief discussion of the research methodology is provided in the third section. In the fourth section, an approach to the estimation of benefit of ITS in Incident Management estimation in this specific context is outlined in greater detail. After presentation of the results in the fifth, some conclusions concerning the estimation of benefits of ITS technology application in incident management are presented. BACKGROUND Each year, traffic congestion leads to millions of hours of vehicle delays and causes significant losses in productivity, increases in fuel consumption and environmental pollution. The monetary cost of traffic congestion to the US economy has increased in recent decades and is expected to rise to $88 billion per year by 2005 (1). Congestion poses a particular challenge to the viability and vitality of urban centers, metropolitan areas, and economic core regions. By 2005, urban freeway delays are expected to reach four billions annually, an increase of well over 400 percent since 1985 (2). A variety of approaches for dealing with congestion have been proposed and are being implemented across the nation. Congestion results when peak hour traffic demand outstrips the available roadway capacity on a regular basis. This type of congestion is predictable. Three general approaches for dealing with recurrent congestion can be distinguished. Capacity expansion is one way to alleviate recurrent congestion. However, when traffic demand and its distribution over the day remain stable this in an expensive solution because the added capacity will be underutilized during off-peak hours. When traffic demand continues to grow then capacity expansion provides only a temporary fix. Experience shows that new capacity tends to be absorbed quickly. Further, many localities face fiscal and spatial constraints, which limit capacity expansion. For these reasons, capacity expansion alone may not be a solution that is sustainable in the long run. Another way to address recurrent congestion is by managing travel demand. Demand management may include the promotion of modes of transportation other than the automobile as well as a variety of pricing schemes. However, influencing travel and transportation-related attitudes and behaviors is a highly complex long-term endeavor that involves many factors, greater uncertainties, and less predictable outcomes. Furthermore, demand management involves sensitive issues of equity that require careful consideration.
Maas 2 In view of the inherent limitations of capacity expansion as a solution to recurrent congestion and the complexities of demand management, assuring optimal use of existing transportation infrastructure in general, and roadway capacity in particular, has become increasingly important. New and emerging technologies, collectively known as Intelligent Transportation Systems [ITS], find application in the context of optimization of transportation infrastructure usage. ITS incorporate a smoothing or rationalizing process that is intended to make the existing supply of transportation infrastructure better meet current and future transportation demand. By employing advanced technologies, the infrastructure is made more efficient, effective and safer while precluding increased land-use and environmental pollution. In addition, ITS serve as a policy facilitator through which public policy can be more effective in encouraging energy saving and discouraging road travel (3). Although ITS is technology oriented in the sense of being directed at improvement of the engineering subsystem, they also comprise information and human subsystems (4). While a wide range of technologies has potential application, the focus of ITS lies with five areas: 1) Advanced Traffic Management Systems, 2) Advanced Traveler Information Systems, 3) Automated Vehicle Control Systems, 4) Commercial Vehicle Operations, and 5) Advanced Public Transportation Systems (3). However, a structural imbalance between travel demand and available roadway-capacity is not the only source of congestion. Traffic accidents and other incidents disrupt normal traffic flow and create temporary reductions in roadway capacity that also to congestion. Due to the more random nature of incident occurrence, this non-recurrent congestion is much less predictable and not limited to peak travel periods. Traffic incidents compound recurrent congestion when they occur during peak periods, and when the location and magnitude of incidents in off-peak periods are such that their aftermath impacts peak period travel conditions. Conversely, some types of incidents (e.g. mechanical breakdowns of vehicles) are more likely to occur under the conditions created by recurrent congestion. Non-recurrent congestion from traffic accidents and other incidents is a major component of the overall congestion problem. It is reported that nearly 60% of traffic delays is caused by non-recurrent congestion (5). The Federal Highway Administration [FWHA] estimates that by 2005 incident related congestion will constitute 70% of all urban freeway congestion, with a associated cost to road users of $35 billion (6). Incident detection, response and clearance are not new activities. However, with increasing levels of urban congestion, the need for a more systematic and integrated approach that transcends jurisdictional boundaries has emerged. Incident management requires coordinated efforts by multiple agencies within and between jurisdictions, as well as non-governmental actors. To mitigate non-recurrent congestion from incidents and to assure optimal availability of existing roadway capacity, a growing number of states and metropolitan areas are initiating Incident Management [IM] programs. The purpose of IM is to quickly detect, verify and clear temporary obstructions on roadways in a particular operational area, to provide information about such obstructions, and to restore normal traffic flow as rapidly as possible. Each phase of the incident management process poses particular technical and operational challenges. A combination of strategies and systems is typically employed. With the emergence of ITS, there are opportunities to increase the timeliness, effectiveness, and efficiency of incident management efforts. The evolution of ITS not only promises new tools for incident detection and verification but also for response vehicle fleet management, traffic diversion, and the provision of traveler information.
Maas 3 The National Incident Management Coalition reports that most IM efforts comprise a patchwork of agencies and activities combined in an ad hoc way. That is, few IM programs are strategically organized, with dedicated funding and clear lines of authority. IM programs are also reported as lacking in visibility to the general public and sometimes being perceived as ineffective. Building a mandate is identified as perhaps the single most important step in developing an incident management program" (7). A basic strategy for building a mandate involves raising awareness and interest of key public and private stakeholders in IM. The level of awareness must reach the point where these stakeholders are ready to make organizational changes or commit resources to a program. Accurate estimates of benefits for IM programs can demonstrate that these are a cost-effective means of dealing with non-recurrent congestion. Traditional cost-benefit analysis methodology can be considered adequate for the task although the fact that many IM programs are new, (ideally) regional in scope and involve "spill-over" benefits and costs may complicate the quantification of the benefits of IM. Since it is the objective of U.S. DOT to see ITS core-technology deployed in 75 major metropolitan areas over the next decade, the explicit linkage of IM to ITS deployment may provide additional momentum in building a mandate and securing resources (8). Although the number of IM programs is growing, attempts to obtain or strengthen funding for the deployment of ITS technology in IM are often stymied. Funding is limited and subject to intense competition from more traditional highway projects (8). State and local government agencies report difficulties in preparing budget justifications and cost-benefit analyses of new IM approaches and technologies. Specifically, agency managers and policy makers are often unable to satisfactorily quantify the benefits of new ITS technology applications in IM. In order to create and sustain effective IM programs, it is essential to develop an accurate means of estimating the benefits of ITS applications in IM. However, the paradigm shift in transportation technology and vision, as motivated by the emergence of ITS, may require a change in evaluation methods (9). PURPOSE AND SCOPE Northern Virginia has its share of traffic and congestion problems. The traffic situation in Northern Virginia is very much determined by its location along I-95 and its proximity to the National Capital. The National Capital region has the dubious distinction of being among the top three metropolitan areas in the US in terms of traffic delays. It also has the highest aggregate cost of delay in the nation. Projections for the National Capital region are similar to those for the nation, if not more pronounced. Urban travel in the region is increasing at an estimated rate of 4% per year (10). Trip lengths are expected to increase while home-to-work car occupancy is expected to decrease (11). By 2010, vehicle demand will outstrip highway capacity, not only on the Capital Beltway and major highways but also throughout the remainder of the area's road system. Vehicle Miles of Travel [VMT] in 2020 is expected to be twice that of 1988 (12). If present trends continue, costly congestion will be prevalent. Indeed, a five-fold increase in freeway congestion is expected between 1989 and 2009 (13). In response to increases in recurrent congestion in normal peak hours, peak hour spreading has occurred. The probability of an incident occurring during the peak period is subsequently greater. Further, the aftermath of non-peak hour incidents may be more likely to impact peak period travel conditions. Even though incident rates are decreasing relative to the amount of travel, the consequences of these accidents and other incidents on the region's congested
Maas 4 freeways are becoming more and more severe due to the increasingly heavy loading of the area's road system. Congestion and costly delays due to incidents are thus likely to become even more prevalent than they are today. A formal incident management program has existed in Virginia since the 1980's, with funding from federal, state and local government sources, as well as from motorist reimbursements. A variety of detection and verification techniques are currently employed, including routine police patrols, courtesy patrols (by police, public sector agencies, and the private sector), cellular phones, commercial traffic reports, automatic detectors, and closed circuit television. Current response and clearance techniques include a formal response procedure, a quick clearance policy, as well as the use of Traffic Operations Centers, Traffic Management Teams, and Incident Management Teams. There are agency owned tow trucks as well as towing rotation agreements and towing contracts. A variety of response and clearance techniques are planned, including an incident command system, an emergency response plan, HAZMAT response, and personnel resource lists. Recovery and information techniques include Highway Advisory Radio, Variable Message Signs, an 800 number for motorist information, a construction news service and media partnerships (8). In view of current trends in urban travel and congestion in the National Capital region, and their impact in Virginia, continuous improvement of incident management efforts is necessary. Further ITS technology deployment is expected in the short to medium run. The research discussed here was conducted for the Virginia Department of Transportation [VDOT]. It had two distinct but related purposes: (1) Provide an approach to the estimation of benefits which can result from various ITS applications in IM, and (2) Generate estimates of the benefits of several ITS enhanced IM scenarios for the Northern Virginia region. Library- and field research, interviews, and simulation modeling were employed to develop a practical approach to the benefit estimation problem at hand. This approach builds on recent efforts at incident prediction and benefit-cost analyses related to alternative IM strategies. It may be used by agencies to better describe and estimate the benefits of ITS enhanced IM scenarios on a regional basis. Estimated delay reduction benefits ITS-enhanced IM scenarios, as well as a preliminary valuation of the delay reduction benefits were generated for main roadway in northern Virginia which represent a substantial portion of the road network, traffic and congestion in the National Capital Area. METHODOLOGY A series of field site visits and interviews with local authorities responsible for or involved in IM was conducted to identify general ITS technology areas and specific applications that were expected to significantly assist local agencies in achieving regional IM benefits. Interviews where conducted with VDOT staff of the Northern Virginia District Office and the Traffic Operations Center in Fairfax, VA, the Traffic Management Center in Arlington, VA, the Intelligent Transportation Systems Office in Richmond, VA). Also interviewed were staff of the University Center for
Maas 5 Transportation Research, Virginia Polytechnic Institute and State University in Blacksburg, VA, the Traffic & Parking Services Division of the Montgomery Country Department of Transportation in Rockville, MD, and Maryland State Highway Administration field offices. The literature review and information obtained during field site visits suggest four ITS technology main areas that hold promise of real-world benefits in the Northern Virginia region. On the basis of information from conducted interviews, four ITS applications were selected (Closed Circuit Television, In-vehicle Cellular Phones, Computer Aided Dispatch Screens, and Global Positioning System location). These applications are expected to yield savings in incident detection-, verification-, response- and/or clearance time and to have a substantial impact on the IM process and subsequent delay reduction. All four applications have a short to medium-term field deployment horizon. In parallel, a comprehension review of the literature on IM, ITS and benefit estimation was conducted to identify state-of-the-art approaches to benefit estimation, suitable for 'What If?' analyses with various possible ITS-augmented IM scenarios. Information was obtained from a variety of sources, including the ITS America, the Texas Transportation Institute, and PATH databases. After exploring several options, the IMPACT v1.0 incident impact modeling software was selected as a most promising benefit estimation tool. Upon adaptation and calibration to local circumstances, the model was used to generate estimates for the Northern Virginia context. BENEFIT ESTIMATION APPROACH Scenario Development The literature review and information obtained during field site visits suggested four main ITS technology areas that hold promise of real-world benefits in the Northern Virginia region: Advanced surveillance and detection applications, Response vehicle communications integration, Routing and scheduling algorithms for response vehicles, and Location applications for response vehicles. On the basis of information from conducted interviews, four specific ITS applications were selected. These were expected to yield the greatest savings in incident detection-, verification-, responseand/or clearance time and thus have a substantial impact on the IM process and subsequent delay reduction. All four applications have a short to medium-term field deployment horizon. Next, estimates of the time saving that each ITS application could generate were required. Since data illustrating the timesaving effect of ITS applications in IM is virtually non-existent, interviews with IM program officials are often the only viable alternative source of information. Local IM expert were asked to provide an indication of the impact they expected from deployment of each of the four ITS applications. Specifically, they were asked to provide an estimate of the reduction of incident duration for all incidents. The four ITS applications and their associated minimum and maximum average incident duration reduction effects are:
Maas 6 Closed Circuit TV (4 to 7 minutes) Cellular Phone in response vehicles (2 to 5 minutes) Computer Aided Dispatch Screens in response vehicles (2 to 5 minutes) Global Positioning System location for response vehicles (4 to 7 minutes). The reduction of the average incident duration for each of the four ITS applications, was given under the assumption of full deployment of single applications in northern Virginia. However, deployment of two or more of the applications simultaneously is likely to result in some overlapping of incident duration reducing effects. As with combinations of IM strategies, the effects of concurrent deployment of ITS applications in IM are cumulative but not necessarily additive. Listed below are the five probably ITS-enhanced IM scenarios that would be explored, each with applicable values for the reduction of average incident duration. CCTV Only: 4 minutes CCTV and Cellular Phone: 6 minutes CCTV, Cellular Phone, and GPS: 9 minutes CCTV, Cellular Phone, GPS, and CAD (minimum effect): 13 minutes CCTV, Cellular Phone, GPS, and CAD (maximum effect): 19 minutes Incident Impact Modeling Software After exploring several options, the IMPACT v1.0 incident impact modeling software was selected as a most promising benefit estimation tool. IMPACT v1.0 is a computer model of incident occurrence, location and severity, that is intended for use in estimating incident impacts for urban freeway segments and traffic volumes, in terms of resulting delays. It is also intended for the quantification of expected changes in the delays from incidents that correspond to alternative improvements to freeways, traffic management, and incident management procedures. Designed for planning and cost-benefit analysis applications, IMPACT v1.0 is a stand-alone computer program implemented in the Windows environment. The program uses standard Windows file management and help screen facilities, and its entirely menu driven through its stages of input data specification, solution calculation, and output of results (14). The software was developed under contract DTFH61-93-C-00015 between the FHWA, U.S. DOT, and Ball Systems Engineering, with the latter subcontracting to the California State Polytechnic University at San Luis Obispo. To generate the model, the developers collected and analyzed the best available incident, highway geometry and traffic volume data from eight major U.S. cities. The statistical analysis revealed considerable similarities in the patterns in incident frequencies and characteristics that were quantified for these cities. Differences among locations could be explained by variations in traffic conditions, incident response capabilities and other known factors. These empirical relationships have been captured in the IMPACT model. In view of the pattern similarities and the fact that known factors could explain what differences where found, the developers of the model argue that many of the empirical relationships that were developed are transferable to other locations (15).
Maas 7 Specifically, the IMPACT model comprises four sets of calculations: Freeway capacity, Incident rates, Incident location and severity, and Incident Duration and Delays due to incidents. These sub-models link together to estimate the impact of each of seven aggregate freeway incident types, under any one of five IM scenarios: No Incident Management, Traffic Management Center [TMC], Major Incident Response Team [MIRT], Freeway Service Patrol [FSP], and User Defined. Calculations are performed on a road section by road section basis, with no spillover effects from one section to the next. Local road section geometry and traffic volumes constitute the minimum required data inputs. As implemented, the software includes default values for most model parameters. Analyses can be performed in the absence of knowledge of specific local values for a majority of the model parameters (15). Comparison of local accident data with the results from trial runs with the IMPACT model suggests that reasonably accurate estimates for the number of incidents can be obtained using these default values. Extensive local calibration of the model does not appear necessary. Other aspects of the model, however, do require adaptation and/or recalibration. The calculation of estimates of delays resulting from incidents is of primary interest in the context of this research. The model estimates annual hours of delay resulting from incidents. Delays on every single road section are estimated for each of seven types of incidents, under any one of five implemented IM scenarios. A summary total is provided for each road section, and for all road sections in the analysis. The user must run the model several times in order to assess the delay reduction effects of various IM scenarios. The 'No Incident Management' scenario must be run to establish a baseline value for delays. The user must also run either one of the three IM scenarios that are implemented in the software (TMC, MIRT, and FSP) or a 'User Defined' scenario. Finally, to arrive at the estimated delay reduction due to a particular IM strategy, the delay estimates for that IM scenario must be related and compared to those under the 'No Incident Management' scenario. But for two aspects, the general set-up and method of use of the model was suitable to the estimation problem at hand. First, the model allows the user to select only a single IM strategy (TMC, MIRT or FSP) while the current IM effort in northern Virginia employs all three in combination. Second, no ITS-enhanced IM scenarios are implemented in the model. In order to use the model for the specific purposes of this research, two adaptations were required. The first was the implementation of a baseline scenario reflecting the current IM effort in northern Virginia. 1 The second adaptation involved the detailing, development and implementation of several ITS-enhanced IM scenarios.
Maas 8 Model Adaptation In the IMPACT model, the number of incidents of each type that was estimated in the incident rate sub-model is multiplied by percentages associated with various lateral locations of incidents on a roadway section obtained from the location and severity sub-model. Then, the number of incidents for each lateral location, together with its characteristic reduction of capacity, is combined with four percentile values that represent the duration distribution of incidents of a given type, which occur in one of two lateral locations, under a given IM scenario. Finally, each combination of capacity reduction and incident duration is evaluated in the delay sub-model. The delay sub-model estimates queue delays based on the cumulative arrival-departure curve method described by Morales (16). The methods employs two curves: 1) a cumulative arrival curve, representing normal traffic demand, and 2) a departure curve which follows the arrival curve if demand is less than capacity, or is determined by capacity when demand exceeds capacity. The slope of the arrival curve is the desired traffic flow rate (in vehicles per hour). That of the departure curve is the value of capacity during times of capacity constraint. The area between the curves is the total vehicle-hours of delay for the period of constrained capacity. Each type of incident with a characteristic severity and duration - is assumed to occur at a number of specific clock times, distributed throughout the peak or off-peak period being considered. These evaluation times have been selected by the developers of the model and provide a representative sample of times-of-day with significantly different delay impact consequences. Four to eight different evaluation times are used, depending on the shape of the arrival traffic demand distribution. Each evaluation time is associated with a weight that represents the proportion of incidents in the period represented by that evaluation time. Given a cumulative arrival curve, the values of normal and reduced capacity associated with the roadway geometry and incident type, and the four percentile values of incident duration associated with the incident type, the vehicle-hours of delay is calculated for each representative incident at each evaluation time by solving the geometry for the area between the curves. 2 The adaptations in the context of this research could be implemented using the 'User Defined' option in the software's Incident Management Strategy dialog box, which is part of the delay sub-model. A 'User Defined' IM scenario may be set up by providing eight new percentile values for each of seven incident types. When the user does not specify any new percentile value, the 'User Defined' option defaults to settings that are identical to those of the 'No Incident Management' option. Following procedures described in the original software development report, the incident duration distribution percentile values can be derived from mean and standard deviation values of incident duration (15). In the absence of valid and reliable local incident duration data, a baseline for the current IM effort in Northern Virginia was constructed by selecting percentile values from among the ones already implemented in the model. That is, the set of model parameters that yielded the greatest reduction in incident duration for each type of incident was selected. The underlying rationale was that, if the three IM strategies were employed concurrently, their delay reducing effects would be cumulative although not additive. It was assumed that this 'optimal' selection from the available incident duration distribution parameters could adequately reflect the combination of traffic management
Maas 9 centers, freeway service patrols, and major incident response teams employed in northern Virginia. The worst eventuality would be that the delay reducing effect of the scenario for the combination of all three IM strategies would be underestimated. Underlying the incident duration distribution percentile values of the newly specified 'Northern Virginia Baseline' IM scenario are mean incident duration values and associated standard deviations for each type of incident, by lateral location of the incidents. Using these means and standard deviations, seven 'percentile value look-up tables' were created. That is, one for each type of incident. These tables show the Baseline mean incident duration, and consecutively decreased mean incident durations in one minute increments, adjusted standard deviations proportional to the means, and the incident duration distribution percentile values associated with each set of one mean and one standard deviation. These tables were employed in implementing the ITS enhanced scenarios that were developed in parallel. Road Section Input Data Local calibration of the IMPACT model required determination of the road sections to be used for estimating the benefits of ITS deployment in IM in northern Virginia. The research focused on nine roadway archetypes, which represent some typical situations that may be found in metropolitan areas. An example of each archetype was identified for the Northern Virginia region. The examples provide complete coverage of the northern Virginia portion of the Capital Beltway (I-495), plus the two main interstate highway routes that feed into Washington DC from Virginia (I-95 and I 66). The portions of the feeder routes inside and outside the main circumferential (Capital Beltway) are considered distinct. The portions of two arterials (Rt-28 and Rt-123) that were selected may be considered secondary circumferentials further away from the main urban concentration. A portion of U.S.-50 was included to serve as a basis of comparison with I-66 and/or I-95 outside the Beltway: these routes all function as spokes in the local hub. To run the model, the necessary data elements for each roadway were collected and/or constructed. The roadway archetypes are sub-divided into a total of 36 smaller sections, based on the heterogeneity of the roadways. For each roadway section, the first data items that were specified are mileage and average annual daily traffic [AADT]. Highway geometry and performance data was obtained from VDOT (18). With respect to traffic and road conditions within the roadway archetypes, several assumptions were made in order to achieve full local calibration of the roadway section data as required by IMPACT. A generalized view of the travel peaks in the Northern Virginia region suggests that the morning peak period starts between 6 a.m. and 7 a.m., and ends between 9 a.m. and 10 a.m. (three hours). The afternoon peak period starts around 2 p.m. and lasts until about 7 p.m. (five hours). Since the average length of trip on the arterials is generally shorter than that on the interstate highways, the length of the total peak period on the arterials is assumed to be one hour shorter. Therefore, the daily peak period length was set at eight hours total for all interstate highway mileage, and at seven hours total for arterial mileage. Values for the peak hour share of ADT ( K-factor in the IMPACT model) were derived on the basis of available data for the Woodrow Wilson Bridge on
Maas 10 the National Capital Beltway, I-95/495, and validated in trial runs with the model. The K- factor value was set at 8% for Interstate highways. For arterials, a slightly lower value of 7% was used, in view of the shorter average lengths of trips on such roads. Similarly it was assumed that, during peak hours, traffic on circumferential routes (e.g. I-495, Rt-28) has no clearly heavier direction. In these cases, the peak hour directional factor D was set at 50%. For through routes or commuter routes, the directional factor D has initially been set at 70% to reflect that in the morning most traffic is in-bound and in de evening most traffic is out-bound relative to the national capital. For some sections, however, the values for the D factor were adjusted adjustment downwards because high D values caused the IMPACT model's limit to the maximum throughput of sections to be exceeded. These limits are based on the assumed average speed and the recommended throughput design of the roads. It is common knowledge that the National Capital region is among the most congested in the country. As a result, the actual throughput on road sections often exceeds throughput design values. The model accommodates such situations by accepting throughput values of up to 15% greater than design throughput, without this resulting in any error in the model s calculations. However, for some road sections, the actual throughput exceeds design throughput by more than 15%. In these cases adjustments must be made in order to avoid error messages and overflow in the results. The IMPACT model allows, although with maximum caution, for the default design throughput values to be changed. However, when changes are made, the model appears to become unstable, producing results that vary to great extremes. Therefore, rather than tampering with these defaults at the risk of model instability and error, minor changes were made in the inputs instead Depending on the truck definition employed, about 12% to 15% of the AADT generally consists of trucks. During peak periods, the percentage trucks drops to about 5% to 7%. In view of the fact that truck drivers tend to avoid congested regions, such as Northern Virginia, during peak periods of the day if they can, the peak period truck shares for IMPACT were set to: 0% on truck embargo routes (e.g. I-66 inside Beltway) 4% on commuter routes 6% on through truck routes (e.g. I-95) Finally, the value for the ratio of average annual weekday traffic to daily traffic (AAWT/AADT) was set at 1.15. This measure of weekday intensity of traffic for this region was based on a downward adjustment of the actual ratio at the Woodrow Wilson Bridge (I-95/495) which is 1.164. This ratio (1.15) may be interpreted to mean that the region experiences 15% more vehicles on an average workday (i.e. Monday through Friday) than on an average day of the entire week. The data on the number of lanes and presence or absence of shoulders on roadways was field-verified for all road sections in the analysis. Next, the data for the 36 stretches of roadways were inputted into the IMPACT model, and the total yearly delay for all these sections calculated under the No Incident Management option. Since the required parameters for the User Defined IM scenarios were derived, the model s management parameters could be varied accordingly and the model run repeatedly with these different sets of incident duration distribution parameter values. The results of seven model runs are summarized and discussed in the Results section.
Maas 11 RESULTS Total Travel, Incidents, and Accidents The IMPACT model was run repeatedly with identical road section data, with each run representing a different IM scenario. Total annual vehicle-miles of travel (VMT) for the 36 road sections in Northern Virginia was estimated at 2,900,000,000, with 55% of VMT occurring during peak travel hours. The total number of incidents on these roadway segments (accidents and vehicle fires, mechanical and electrical breakdowns, dropped loads and debris on the road, vehicle stalls, flat tires, and abandoned vehicles) was estimated to be about 34,000 annually, with 77% of the incidents occurring during peak hours. Of this, the total number of accidents and vehicle fires was estimated to be 3200, or about 9% of the total number of incidents. These figures were constant throughout all seven model-runs. That is, only the estimates for the delays resulting from these incidents varied. Incident Impacts and Delay Reductions Achieved through Current Incident Management. Without IM, the incidents mentioned previously would result in an estimated 4,242,000 hours of delay annually. In monetary terms, with one hour of delay uniformly valued at $10, this is roughly equivalent to a loss to the region of $42 million. Delays due to the incidents dropped to 2,786,000 hours per year under the scenario representing the current IM efforts in Northern Virginia, that is a combination of TMCs, MIRTs and FSP. In other words, the delays on the 36 road segments alone are reduced by 1,456,000 hours annually (34%), which may be conservatively valued at $14.5 millions. The delays are most likely costlier than this each year, if all road segments are added in, and higher time valuation is included for trucks and freight, for example. By assuming a $10 time-value, a conservative estimate of monetary savings is provided. Heavy truck delay costs for driver, vehicle, and freight falls anywhere in the range from $25 to $100 per hour. Thus, given 5-7% trucks on the roadways, actual savings may be greater than reported here. Accidents and vehicle fires as a share (9%) of total incidents conform to the national average. (10). For the northern Virginia region, however, it is possible that the model underestimates the share of accidents and vehicle fires. This particular region has one of the highest per-capita income levels in the nation and, correspondingly, one of the newest and most expensive automobile fleets. One would expect fewer mechanical breakdowns, flat tires, stalls, and abandoned vehicles with a fleet of this type. If, in reality, accidents and vehicle fires were to constitute a larger proportion of total incidents, one would expect more congestion and more delay. Since the 9% share seems low for the region, the number of hours of delay presented here may be conservative. In view of the relatively simple method of valuation that was applied, the delay cost-estimates are most certainly conservative.
Maas 12 Estimated Benefits from ITS Deployment in Incident Management. Five specific ITS-enhanced IM scenarios were examined. Since concurrent deployment of all four ITS technologies and subsequent realization of their maximum estimated benefits (ITSenhanced incident management scenarios 4 and 5) may be somewhat optimistic in the short to medium run, only the results of first three scenarios are discussed in some detail. These scenarios appear very realistic within the given time horizon. The results of the modeling runs for the first three ITS-enhanced IM scenarios were compared with those of both the No Incident Management scenario and the scenario reflecting the current IM capability in northern Virginia. Thus, estimates of additional benefits of (combinations of) ITS applications were obtained. With the current IM effort augmented by full deployment of CCTV (scenario 1), annual hours of delay on the selected segments in Northern Virginia dropped further to 2,192,000. As compared to no IM, this is a reduction of 48%. In this scenario, the ITS augmentation results in a 21% gain over the current IM effort. The additional benefits from deploying CCTV for IM is worth about $6 million and would raise the total annual savings as a result of IM in Northern Virginia on the selected segments to $20.5 million. Using a combination of CCTV and Cellular phones in response vehicles to augment current IM efforts (scenario 2), annual delays on the roadway segments dropped to 1,797,000 hours. That is a reduction of 58% compared to having no IM activity whatsoever, and an improvement of 35% over the region's current IM effort. This ITS-enhanced IM scenario is worth about $10 million and raises the total annual savings due to incident management to as much as $24.5 million for the Northern Virginia road segments that were analyzed. The third ITS-enhanced IM scenario, which adds in-vehicle GPS location to CCTV and cellular phone based communications, would reduce the annual hours of delay down to 1,515,000. Compared to having no IM capability in place, this is a delay reduction of 64%. A delay reduction of 46% over the current IM effort would be achieved. The total annual savings due to IM increase to $27.3 million for the selected road segments; the additional benefits from the third ITS-enhancement scenario is worth $12.5 million in savings. Some caution is warranted. The delay estimates for the northern Virginia region are considered about as accurate as they could be, in view of the current implementation of the IMPACT model. Nevertheless, there are some features of both the model and the region that require careful consideration when interpreting the results of this study. The model calculates highway segment capacity on the basis of procedures and default values taken from the Highway Capacity Manual (19). These values can be modified, but only within a limited specified range. The calculated section capacities have considerable influence on the subsequent calculations of incident delays according to the manual for the modeling software (9). These procedures and default values may play a role in underestimating highway segment capacity as it exists in the northern Virginia Region. For example, in this region, actual highway segment capacities may be greater that assumed in the model due to several factors, which are expected to become increasingly characteristic of the region s highway network in the future:
Maas 13 Shoulder lanes opened for travel during peak hours Dedicated express lanes open for travel during peak hours reversible HOV lanes Furthermore, actual traffic volumes in this region are significantly in excess of model assumptions; weekday peak volumes on Interstates are typically 150% to 200% of design capacity. The region has been able to achieve these extraordinarily high volumes through a combination of high rush hour speeds, dangerously close vehicle spacing, and round-the-clock driving. Most importantly, this region exhibits significant peak spreading when compared to other regions. That is, traffic volume builds earlier and continues longer in both the morning and afternoon peak periods, making the peak periods longer while enabling these higher volumes to move through the highway network. Therefore, in view of the freeway capacity-enhancing features that are employed in this region, and the subsequently greater-than-assumed traffic volumes, the model in its current form probably underestimates true highway capacity. Hence, the traffic volumes accepted by the model, the estimated number of incidents, and the resulting incident delays presented here are likely to be less than what they are in reality. The true delay and delay cost estimates in northern Virginia are probably greater than the estimates developed for the roadway segments that were tested. Another consideration is that while the approach developed here allows separating out the added effects of ITS, it essentially results in a snapshot. That is, the benefits of ITS in IM were calculated on the basis of traffic and capacity in a given year. Given the short-to-medium range deployment horizon of the ITS applications analyzed here, roadway capacity may be considered. However, an increase in travel with a concomitant increase in the number of incidents or more severe incident impacts will off set some of the benefits. To assess the effect of changes in the amount of travel, the current analysis could be repeated with data for several years. For long deployment horizons, a dynamic model is probably more suited. Finally, the analysis focussed on estimating the number of hours of delay and the delay reduction as, respectively, the cost of congestion and the benefits of IM and ITS. An initial valuation of these costs and benefits was provided. However, this valuation serves only as a general indicator for the magnitude of the monetary aspects involved, and not as a definite financial estimate. CONCLUSIONS Although the IMPACT model has some shortcomings, it is based on some of the best available data from significant sources of experience with IM in the country. Many assumptions were build into the model by its developer, and many assumptions were added in this research due to lack of precise information. However, as the developers of IMPACT have realized, there are valid statistical trends that can be employed sensibly. It is concluded, therefore, that while the IMPACT model can be upgraded in a variety of ways, it is among the best currently implemented. Furthermore, the IMPACT model can be used to provide the benefit estimations not only of
Maas 14 various IM strategies independently but - with some re-calibration and adaptation - also of combinations of IM strategies and ITS-enhanced IM strategies. It can play a significant role in planning, justifying, and funding continued incident management efforts and the deployment of ITS in incident management. With respect to the estimation of the benefits of ITS applications in IM, the research report illustrates that ITS technology provides the greatest impact in the early stages of the incident management process, and therefore has an immense effect on the yearly hours of delay saved. In Northern Virginia, an estimated 35% reduction of incident delays is already being realized through currently deployed IM strategies and systems. It estimated that full deployment of selected applications of ITS technology on roadways, traffic management centers, and in response vehicles can reduce current weekday delays by 21 to 46 percent. All other things remaining equal, ITS deployment could double the effectiveness of the current IM effort. ACKNOWLEDGEMENTS The research project "Benefits Evaluation of Incident Response and ITS: Accidents and Nonrecurrent Congestion in the National Capital Region" was conducted with sponsorship from the Virginia Transportation Research Council for the Virginia Department of Transportation. The author wishes to acknowledge the support of both VDOT and the VTRC in making this project possible. The project started on 03-15-1996 and was completed on 06-30-1997 with the delivery of the report entitled "Final Report: Application of the IMPACT Model to ITS Benefit Estimation". The research for this project was conducted by The Institute of Public Policy at George Mason University, Fairfax VA. The Principal Investigator for the project was Roger R. Stough, Director of the Transportation Policy Program. The core research staff consisted of Gerard C. Maas, Mark E. Maggio, and Hadi Shafie, with support from Paul M.A. Baker, Lauri Schintler, and Raj Kulkarni. The final report can be obtained from The Institute of Public Policy. REFERENCES 1 Institute of Transportation Engineers. Mobility Facts, Washington DC, 1996. 2 U. S. Department of Transportation. Federal Highway Administration. Highway Statistics 1987. Washington, DC: US-GPO, 1987. 3 Haynes, K. E., F. Y. Phillips, L. Qiangsheng, N. Pandit, & C. R. Arieira. Information Based Uncertainty Management of IVHS/GIS-T Technologies. Paper presented at the International Symposium on IVHS/GIS-T, Seoul, South Korea, June 1994. 4 Padgett, R. L., Human Error and Risk Management Approaches to Cost-Effectiveness and Decisionmaking in Intelligent Transportation System Development and Evaluation. Dissertation submitted for the degree of Doctor of Philosophy in Information Technology, Fairfax VA: George Mason University, 1996. 5 Cambridge Systematics Inc. Incident Management. Final Report Prepared for Trucking Research Institute. ATA Foundation Inc. Alexandria VA, 1990. 6 Mannering, F. L., M. Hallenbeck, and J. Koehne. A Framework for Developing Incident Mangement Systems: A Summary. Report WA-RD 224.2. Washington State Transportation
Maas 15 Center, University of Washington, 1992. 7 Cambridge Systematics Inc. Incident Management. Challenges, Strategies, and Solutions for Advancing Safety and Road Way Efficiency. Executive Summary Prepared for Trucking Research Institute. ATA Foundation Inc. Alexandria VA, 1997, p.12. 8 Cambridge Systematics Inc. Incident Management. Challenges, Strategies, and Solutions for Advancing Safety and Road Way Efficiency. Full Report Prepared for Trucking Research Institute. ATA Foundation Inc. Alexandria VA, 1997. 9 Jin, D. J., and R. R. Stough. Paradigms and ITS Evaluation and Deployment. Paper Submitted to the Transportation Research Board for Presentation at the 1997 Annual Conference. The Institute of Public Policy, Fairfax VA, 1996. 10 Lo, H. K., A. Chatterjee, F. Wegmann, S. Roberts, and A. K. Rathi. Evaluation Framework for IVHS. Journal of Transportation Engineering, Vol. 120, No. 3, May, pp. 447-460. 11 Greater Washington Transportation Planning Board. The Highway and Transit Facility Element of the Long-Range Transportation Plan for the National Capital Region. Washington DC, December 1990. 12 Mobility 2000. Final Report of the Working Group on Operational Benefits. Mobility 2000, San Antonio TX, March 1990. 13 Mobility 2000, Proceedings of a Workshop on Intelligent Vehicle Highway Systems. Mobility 2000, San Antonio TX, February15-17 1990. 14 Sullivan, E., and D. Champion. IMPACT 1.0: A model for the estimation of Incident Impacts on Freeways (User's Guide). FWHA, U.S. Department of Transportation, 1995. 15 Sullivan, E., S. Taff, and J. Daly. Final Report for Tasks H-J Incident Detection Issues: a Methodology for Measurement and Reporting of Incidents and the Prediction of Incident Impacts on Freeways. DTFH61-93-C-00015. FWHA, U.S. Department of Transportation, 1995 16 Morales, J. M. "Analytical Procedures for Estimating Freeway Traffic Congestion". ITE Journal. Institute of Transportation Engineers, Washington DC, January 1987. 17 Lindley, J. A. "A Methodology for Quantifying Urban Freeway Congestion" Transportation Research Record 1132. Transportation Research Board, Washington DC, 1987. 18 Commonwealth of Virginia. Average Daily Traffic Volumes on Interstate, Arterial and Primary Routes. 1995. Virginia Department of Transportation, 1997. 19 Federal Highway Administration. Highway Capacity Manual. 1985. FWHA, U.S. Department of Transportation, 1985. NOTES 1 Although the estimation of the benefit of current IM efforts is not the objective of the research, the baseline scenario reflecting the current IM effort must be developed in order to determine the marginal benefits of ITS applications. That is, to establish the increase in IM benefits due to ITS applications rather than the benefits of current IM efforts per se. 2 The model employs time-of-day traffic volume distributions derived from those employed in the FREWAY model developed by Lindley (17). The shape of these curves is characterized by different values of the product of the percent of weekday ADT in the typical peak hour (K), and the percent of total peak hour traffic in the heavier direction (D). The curves are shifted to correspond to one of