Application of EMME/2 and Enif for a Congestion Relief Analysis Study in the Puget Sound Region

Transcription

1 Application of EMME/2 and Enif for a Congestion Relief Analysis Study in the Puget Sound Region BY Sujay Davuluri Parsons Brinckerhoff Quade & Douglas, Inc. 999 Third Avenue, Suite 2200, Seattle, WA 98104, USA Tel: (206) , Fax: (206) davuluri@pbworld.com Nicola Longo, P.E. Parsons Brinckerhoff Quade & Douglas, Inc. 999 Third Avenue, Suite 2200, Seattle, WA 98104, USA Tel: (206) , Fax: (206) longo@pbworld.com Murli K. Adury, P.E. Parsons Brinckerhoff Quade & Douglas, Inc. 999 Third Avenue, Suite 2200, Seattle, WA 98104, USA Tel: (206) , Fax: (206) adury@pbworld.com and Clyde Scott Parametrix Inc. 411, 108th Avenue NE, Suite 1800, Bellevue, WA Ph.: (425) , Fax: (425) cscott@parametrix.com October 12, 2004 Paper Submitted for Presentation at the 18 th International EMME/2 User s Group Conference Mexico City, Mexico, October 20-22, 2004

2 Application of EMME/2 and Enif for a Congestion Relief Analysis Study in the Puget Sound Region ABSTRACT This paper describes the application of EMME/2 and Enif modeling techniques used for the Congestion Relief Analysis (CRA) study in the Puget Sound Region. The primary purpose of the CRA study was to evaluate a system of transportation improvements to alleviate congestion in the three major urban areas of Washington State: the Puget Sound Region, the City of Vancouver and the City of Spokane. The Puget Sound Region is the largest and most congested urban area in the State of Washington and is the focus for this paper. Forecasts indicate that population and employment growth in the region, would result in increased traffic congestion in the region by the study year The CRA study is aimed at analyzing the effectiveness of various multi-modal alternatives in relieving congestion in the Puget Sound Region. As a first step, two unconstrained scenarios and a pricing scenario were studied to assess the full potential travel demand for highway and transit modes. The highway and transit unconstrained scenarios along with pricing were used to form the basis for defining the other mixed scenarios in the study. EMME/2 related special techniques were required to perform the modeling analysis for the Unconstrained Highway, Unconstrained Transit and the Pricing alternatives. Enif - a complementary tool to EMME/2 was extensively used to create graphics illustrating the various performance indicators. Enif was used as a network analysis tool for coding and as an evaluation analysis tool to compare the alternatives and also for communicating the network assumptions, model results with the study team members and representatives from concerned jurisdictions. The results from the Enif analysis were used later on to develop other alternatives, which looked at various possible combinations of highway and transit capacity additions along with congestion pricing. A variety of performance indicators depicted using Enif, were considered for evaluating the potential alternatives. This paper primarily describes the methodology and procedures used to model the unconstrained highway and transit alternatives. It also describes the methodology used to model pricing based on Dr. Randall Pozdena s augmented volume delay function approach. The paper also describes the use of Enif for this study and highlights how this software was used to analyze, compare and summarize the various alternatives.

3 Davuluri, Longo, Adury, Scott 1 INTRODUCTION The Congestion Relief Analysis (CRA) study s purpose is to evaluate a system of transportation improvements to alleviate congestion in three urban areas of Washington State: the Puget Sound Region, Vancouver, and Spokane. This study is a systems-level planning analysis of a set of transportation improvements for automobile, high-occupancy vehicle (HOV) and transit modes as well as transportation pricing strategies for relieving congestion. The objective of the study is to look at various alternatives that combine multi-modal improvements to alleviate congestion in these urban areas. The results from these various study alternatives will be used as guidelines to provide insights for policy makers to refine investment choices when major new investment funds become available. Puget Sound region is the largest and most congested urban area in the State of Washington and is the primary focus of this paper. Figure 1 shows a map of the Puget Sound Region along with the region s major freeways. Figure 2 identifies some of the key activity centers in the region. Capacity expansion in Puget Sound region has not kept pace with the rapid growth in travel demand. Most existing highways run at nearly congested conditions during peak periods of traffic. Rapid growth in the region is expected to continue, thereby increasing the traffic demand on its already inadequate transportation system. Growth trends for the region show that by the future analysis year 2025, the total population and employment are expected to grow by about 35% and 40% respectively. This growth in population and employment would translate to additional trips in the region and would result in approximately a 50% increase in total vehicle trips as well as vehicle miles traveled (VMT) by the year 2025 when compared against current conditions. The region s proposed highway capacity expansion by 2025 (based on committed projects) would amount to a mere 2% in added capacity for freeways and approximately a 1% capacity increase for other roadways. With most of the current capacity already consumed and the future capacity expansion not on par with the future demand, the region is likely to experience significant congestion levels by This increase in travel demand associated with a minimal increase in roadway capacity will result in roughly 30-fold increase in total vehicle hours of delay regionally. This imbalance between demand and capacity in the region is the primary impetus for this study. This paper describes the application of EMME/2 modeling techniques and the usage of Enif graphic capabilities for the CRA study. A major portion of the paper describes the EMME/2 related modeling methodologies used to model the unconstrained highway and transit scenarios and the congestion-pricing scenario. The paper also discusses the application of Enif for the study and in particular its usage in defining future study alternatives using the results obtained from the unconstrained and pricing scenarios.

4 Davuluri, Longo, Adury, Scott 2 Figure 1 Major Freeways in the Puget Sound Region

5 Davuluri, Longo, Adury, Scott 3 Figure 2 Some of the Key Activity Centers in the Puget Sound Region

6 Davuluri, Longo, Adury, Scott 4 EMME/2 MODELING METHODOLOGIES As described earlier various alternatives that included possible combinations of highway & transit investments along with congestion pricing were considered in the study. Prior to defining the study alternatives, the maximum potential travel demand for highway and transit modes was determined by running capacity unconstrained highway and transit scenarios that required the use of special EMME/2 modeling techniques. In addition to modeling the capacity unconstrained scenarios, the effect of region wide congestion pricing was determined. A pricing model that implemented tolls on all the highway facilities in the region was modeled by using the augmented volume-delay function approach, developed by Dr. Randall Pozdena. The results from the above mentioned capacity unconstrained and pricing scenarios were used to define the final study alternatives. A major portion of this paper primarily concentrates on EMME/2 modeling methodologies used for the capacity unconstrained and the pricing scenarios. APPLICATION OF ENIF Enif - the complementary graphic tool to EMME/2 was extensively used at every stage in the study. Enif proved to be extremely useful for the study as a network analysis tool and as an evaluation tool. As a network analysis tool, Enif was very useful in coding the highway and transit networks for the study. Enif was used to identify and correct network-coding inconsistencies in the highway and transit networks. Enif proved to be very useful in comparing the highway and transit networks among the various study alternatives. Transit network coding was much easier and faster using the transit network plotting capabilities of Enif. Defining the various studying alternatives was an iterative process and network definition plots produced using Enif were very helpful in redefining the alternatives during the iterative process. Enif was also used as an evaluation tool to illustrate variety of results and performance indicators from the modeling analysis of all the study alternatives. Enif was used to compare the results from various alternatives and was used to determine the effectiveness of the alternatives for relieving congestion. Results from Enif analysis proved to be useful in developing new alternatives, and in redefining, modifying some of the existing alternatives. Plots produced from Enif proved to be an easier and effective way to communicate the network assumptions and the model results with the study members and representatives from concerned jurisdictions. Along with describing the modeling methodologies for the unconstrained and pricing models, this paper also describes how Enif was used to display some of the results from these models as well as to define new alternatives based on the results from these models.

7 Davuluri, Longo, Adury, Scott 5 MODELING PROCEDURES: Model Used The travel forecasting analysis for this study was based on the Puget Sound Regional Council s PSRC s) regional model. The PSRC model, developed using EMME/2 is a four-step multi-modal modeling system that performs the traditional four-step process of trip generation, trip distribution, modal choice and trip assignment. The PSRC model was updated and validated to 1998 conditions. It includes several feedback loops based on intra-regional accessibility and interfaces with a land use allocation model. No-Action Model Run The 2025 No-Action alternative is a future do-nothing alternative. Network definition for this alternative was based on the existing conditions network with the inclusion of future committed projects both for highway and transit. The purpose of 2025 No Action model run was to serve as a baseline for comparing other capacity improvement alternatives. The model was also used to develop constrained trip tables and constrained travel times that would be inputs to the unconstrained and pricing models. The 2025 No Action alternative was modeled using the standard four-step modeling process involving feedback loops using constrained highway and transit networks. The word constrained is used to indicates that the standard procedures established for trip production, mode choice and trip assignment in the No-Action takes into account delays due to limited capacity on a roadway (in case of highways) and delays associated with the use of transit, such as accessibility, waiting and transfer. The primary outputs from this model are the constrained person trip table, total daily vehicle trip table, constrained highway and transit travel times. These outputs have been used as inputs for the unconstrained and pricing model runs. A flow chart representing the general methodology for the 2025 No Action alternative is shown in Figure 3. Application of Enif for the No-Action model run Enif was used to display several network assumptions and results from this model and also to compare them with the network assumptions and results of other alternatives. Figures 4 and 5 show the highway and transit networks assumed under the No-Action alternative. Figure 4 shows the number of lanes for the major highway facilities in the region and Figure 5 shows the transit frequency under No-Action conditions in These network plots have been used as the basis for defining highway and transit networks in other alternatives. In a similar fashion plots produced using results from No-Action were used to compare the results from other alternatives. A few plots that compare and analyze the results from this alternative to the other alternatives have been shown later in the paper. Several other output plots based on certain performance indicators were also produced for the No-Action alternative using Enif. A comparison of these

8 Davuluri, Longo, Adury, Scott 6 output plots with similar kind of output plots developed for other study alternatives gave an idea about the performance and effectiveness of various study alternatives. Figure 3 Modeling Methodology for No-Action Model SCHEMATIC FLOW CHART NO-ACTION MODEL Latest 2025 Land Use Forecasts Trip Generation Model: -Trip Ends by Trip Purpose (P s & A s) Trip Distribution Model (Constrained) Constrained Person Trip Tables (MF cptt ) Parking Costs Transit Fare - Other Usual Input Data Mode Choice Model Total Daily Vehicle Trip Table (MF vtt ) Highway Assignment (Capacity Constrained) Constrained Highway Travel Times (MF cht ) Transit Assignment Constrained Transit Travel Times (MF ctt )

9 Davuluri, Longo, Adury, Scott 7 Figure 4 Number of Lanes (Each Direction) for Major Highway Facilities under No-Action Alternative

10 Davuluri, Longo, Adury, Scott 8 Figure 5 Peak Period Transit Service under No-Action Alternative

11 Davuluri, Longo, Adury, Scott 9 Unconstrained Highway Model The Unconstrained Highway Model run was performed to examine the maximum possible auto demand on the highway system. This scenario analyzed the travel patterns when capacity on the roads was unconstrained i.e. there were no congestion related delays in the system. In order to find the true maximum auto demand, transit operations were still considered to be under constrained conditions i.e. similar to the No-Action conditions. A full model run including trip distribution, mode choice and trip assignment was performed to model the effect of shift in trip distribution and mode shift. No feedback loop was required in modeling this scenario as equilibrium conditions are reached in the very first iteration as the path choice and the travel times remain unchanged due to unlimited capacity conditions. Due to the availability of unlimited capacity and free flow conditions everywhere significant increases in trip length are observed. The methodology for this model is presented below. Methodology for Unconstrained Highway Model Step 1 Capacity Unconstrained All or Nothing Highway Assignment: The first step in this model was to assign the total daily vehicle trip table from No Action model on to the capacity unconstrained highway network using an all-or-nothing assignment. Since the network is unconstrained any vehicle trip table can be used for the all-or-nothing assignment, the No-Action total daily vehicle trip table was used, as it was readily available. The assignment was performed by assuming free flow speed on each link i.e., the speed on any given road is independent of the number of vehicles on the link. Free flow speed on the links was modeled by modifying the Volume Delay Function (VDF) to reflect free flow time on the links i.e. VDF = t o, where t o is the time corresponding to free flow speed on the links. The first iteration of the EMME/2 auto assignment is an all-or-nothing assignment and it ensures that all the trips between a zone pair choose the same shortest path based on travel time. Zone to zone unconstrained highway travel times were obtained from this all-or-nothing assignment. Step 2 Trip Distribution: The trip distribution model used the same trip ends (productions and attractions) that have been used in the No-Action model run. The unconstrained highway travel times obtained from the Step 1 were used as the other input into trip distribution model. The result of the trip distribution process was the new trip tables that were distributed based on the unconstrained capacity conditions. Step 3 Mode Choice: The inputs into the mode choice process for this model run were the unconstrained highway travel times obtained in Step 1, constrained transit travel times from the No-Action model, unconstrained demand obtained from trip distribution mentioned above and the other usual inputs

12 Davuluri, Longo, Adury, Scott 10 such as parking costs and transit fares used in No-Action. The output from the mode choice model was the unconstrained total daily vehicle trip tables. Step 4 Capacity Unconstrained All or Nothing Highway Assignment: The total daily vehicle trip table obtained from mode choice (Step 3) was assigned to the capacity unconstrained highway network. Trips were assigned to the highway network using an all-ornothing assignment process using VDF=t o on all the links. This step is similar to Step 1. The primary difference apart from the auto vehicle trip tables, between the two steps was that in Step 1, all-or-nothing assignment was performed to obtain unconstrained highway travel times to be input into trip distribution and mode choice, while in the current step an all-or-nothing assignment was performed to obtain link volumes. The flow chart in Figure 6 depicts the general methodology used for the Unconstrained Highway Model. Application of Enif for the Unconstrained Highway model run The results from the unconstrained highway model were analyzed using Enif and formed the basis for the definition of several highway capacity improvement oriented alternatives. Using the results from this model several plots were developed with Enif. An analysis of these plots indicated the serious need for the additional highway capacity on all the major freeways in the region. Figure 7 shows a volume difference plot between Unconstrained Highway scenario daily vehicle volumes and the No-Action daily vehicle volumes. The figure displays the huge additional demand on the freeway system under unconstrained capacity conditions. By using the information in this plot along with other outputs from the model run, the number of additional lanes required (compared to No-Action) on the major corridors to meet the unconstrained demand, were determined. For a given roadway, additional lanes required were determined by dividing the PM peak hour volume difference between the unconstrained highway model and No- Action for that roadway by its capacity. Since the highway assignment was based an all-ornothing assignment, most of the volumes were assigned to the freeways and there was an underutilization of the arterial capacity in the system (this can be observed in Figure 7). To account for this a screenline approach was used to determine the required additional lanes in the final study alternatives. Once the highway lane additions for the study alternatives were determined, plots similar to Figure 4 showing the lane additions (compared to No-Action) and the total lanes were produced for each alternative.

13 Davuluri, Longo, Adury, Scott 11 Figure 6 Modeling Methodology for Unconstrained Highway Model SCHEMATIC FLOW CHART UNCONSTRAINED HIGHWAY MODEL Use Trip Ends (P s & A s) From No-Action Total Daily Vehicle Trip Table From No-Action (MF vtt ) Unconstrained Highway Travel Times All-or-Nothing Highway Assignment (Capacity Unconstrained) Trip Distribution Model (Unconstrained) Unconstrained Person Trip Tables (MF uptt ) Parking Costs Transit Fare - Other Usual Input Data Mode Choice Model Constrained Transit Travel Times from No-Action (MF ctt ) Create Unconstrained Total Daily Vehicle Trip Table (MF uvtt ) All-or-Nothing Highway Assignment (Capacity Unconstrained) Summarize Results Notes: Capacity Unconstrained highway assignment was based on the using: 1. Free flow speed value on each link, i.e., VDF = t o 2. All-or-Nothing assignment ( no iteration) 3. No feedback through the model chains is necessary

14 Davuluri, Longo, Adury, Scott 12 Figure 7 Difference in Daily Vehicle Volumes (Unconstrained Highway Scenario minus No-Action)

15 Davuluri, Longo, Adury, Scott 13 Unconstrained Transit Model While the unconstrained highway models aimed at forecasting the maximum potential highway demand under constrained transit conditions, the purpose of the unconstrained transit model was to evaluate the maximum potential transit travel demand under constrained conditions for auto vehicles. For this analysis, unconstrained transit conditions meant the existence of a ubiquitous transit network. This meant that transit was accessible and available between all zone pairs, with reduced waiting times and no transfer would be required. In other words every one has his/her own bus. Special procedures were developed to simulate such a ubiquitous transit network, since the manual creation of such a transit network would be problematic and time consuming. Transit assignment in EMME/2 is based on optimal strategies that consider various factors that include transit headways, transit speeds, transit accessibility etc. and ultimately choose the shortest transit path with respect to time. In the ubiquitous transit scenario transit network was present on every highway link in the system and all the transit parameters like headways, speed 1, accessibility etc. were assumed to be constant through out the network. Thus, in this ubiquitous transit network, between a given zone pair there would be only one shortest transit path and that would be the shortest path between that zone pair with respect to distance. An all-or-nothing highway assignment on a highway network with unlimited capacity and a constant speed through out the network would result in the same shortest path between a given zone pair. Hence, the transit demand was assigned to the highway network to replicate ubiquitous transit service. A complete model run was performed, as people are likely to change their destination and shift their mode because of unlimited transit availability. Unlike the unconstrained highway models that involved no feedback loops, the unconstrained transit model was modeled using a feed back loop. This was because the highway network was assumed to operate under constrained conditions for auto vehicles and hence several feedback loops are required for reaching equilibrium. Certain basic assumptions had to be made for modeling the ubiquitous Unconstrained Transit scenario. These basic assumptions include the following: Between any given zone pair the average transit speed is 18 mph on all the links. This average speed was determined based on the existing system wide average transit operating speeds. Walk time from bus stop to origin and from bus stop to destination is 2.5 minutes each Average frequency is 10 minutes in the peak and 15 minutes in off-peak. This translates to a wait time of 5 minutes in the peak and 7.5 minutes in off-peak Transit fares remain unchanged i.e. the transit fares are the same as in No-Action 1 An average transit speed of 18mph was assumed on all the links. Please refer to the first assumption

16 Davuluri, Longo, Adury, Scott 14 Transit transfer time is zero i.e. no transfers are required Weights assumed on transit walk time and wait time are 1.0 Transit accessibility for all the zones is 100% i.e. all the zones have complete access to transit The travel time between a zone pair was assumed to be the minimum of the total travel time calculated using the assumptions above and the No-Action total travel time Described below is the modeling methodology for this scenario that incorporates the above assumptions. Methodology for Unconstrained Transit Model Step 1 Capacity Unconstrained All or Nothing Highway Assignment: The primary purpose of this assignment was to create a transit in-vehicle travel time matrix reflecting an average transit speed of 18 mph on all the links. The daily vehicle trip table obtained from the Unconstrained Highway model was used to perform an unconstrained all-or-nothing highway assignment. For performing this assignment a special volume delay function (VDF) was defined for all the links in the network to reflect the assumed 18 mph transit speed on all the links. The output from this step is the in-vehicle transit travel time. Step 2 Calculation of Total Transit Travel Time: The purpose of this step was to create the total transit travel time matrix to be input into the mode choice process. Total transit time was calculated by adding the total walk time and wait time based on the assumptions stated earlier, to the in-vehicle time obtained from the previous step. The walk time added was 5 minutes for each zone pair, reflecting a walk time of 2.5 minutes at each end of the trip. Similarly the wait time added was 5.0 minutes for peak and 7.5 minutes for off-peak. Once the total transit travel time matrix was obtained, it was compared with the constrained travel time matrix from No-Action, and the minimum travel time between either of them was taken. This step was to ensure that the total transit travel time matrix reflects a travel time that is the minimum possible between a zone pair under the assumed conditions. Step 3 Trip Distribution: As stated earlier this model involved several feedback loops for trip distribution and mode choice. A standard trip distribution similar to the No-Action involving feed back loops was performed. The inputs into the trip distribution were the trips ends (productions and attractions) from No-Action and the highway travel times from the highway assignment in Step 6. The output from trip distribution is the persons trip table.

17 Davuluri, Longo, Adury, Scott 15 Step 4 Mode Choice: A standard mode choice model involving feedback loops was performed. The mode choice process for this model uses the total transit travel times from Step 2, trip tables obtained from Step 3, constrained highway travel times from the highway assignment in Step 6 and the other usual inputs such as parking costs and transit fares used in No-Action. The outputs from the mode choice model are the total daily transit person trip table and the total daily vehicle trip table. Step 5 Capacity Unconstrained All or Nothing Highway Assignment Using Transit Demand: The assignment of the total daily transit person trip table to the network needs to reflect a ubiquitous transit network condition. Based on the discussion earlier in the paper the condition of ubiquitous transit was modeled using an all-or-nothing highway capacity unconstrained assignment for the transit trip tables assuming a constant speed of 18 mph on all the links. The result of such an assignment would be transit passenger volumes on links reflecting the travel patterns and route choices that would have been obtained from a true ubiquitous transit assignment. Step 6 Capacity Constrained Highway Assignments: The total vehicle trip table obtained from the mode choice was assigned to a capacity constrained highway network. The output from these assignments was the constrained highway travel time which was used in Trip Distribution and Mode Choice (Steps 3 & 4) during the feedback loop. The flow chart in Figure 8 depicts the general methodology used for Unconstrained Transit Model. Application of Enif for the Unconstrained Transit model run Results from this model run helped in understanding the potential transit demand in this region. Plots produced from this model run using Enif proved to be extremely useful in defining transit oriented alternatives for the study. Figure 9 compares the No-Action transit volumes with the unconstrained transit volumes. Volumes are shown in two different layers. The top layer in green shows the No-Action Transit volumes while the bottom layer in red shows the Unconstrained Transit Volumes. This plot indicated the need for increase in transit capacity in the region. The plot helped identify the corridors with a need for better transit service. Using the information in this plot and other outputs from the model run, zones and corridors having high potential transit demand were identified and were served accordingly with increased transit capacity/service in the final study alternatives. Once the transit capacity improvements for the study alternatives were determined, plots similar to Figure 5 showing the transit service levels for these alternatives were created.

18 Davuluri, Longo, Adury, Scott 16 Figure 8 Modeling Methodology for Unconstrained Transit Model SCHEMATIC FLOW CHART UNCONSTRAINED TRANSIT MODEL Unconstrained Total Daily Vehicle Trip Table from Model B (MF uvtt ) All-or-Nothing Highway Assignment based on using: VDF = (Link Distance in miles)/spd*60). Save for Transit Travel Times (MF utt ) in minutes Use Trip Ends (P s & A s) From No-Action Calculate Transit Travel Times in minutes: MF utt (Total) = MF utt + Walk Time (MF walk ) + Wait Time (MF wait ) Trip Distribution Model (Constrained) Constrained Total Transit Travel Times from No-Action (MF ctt ) MF tt = min (MF ctt vs. MF utt ) - Walk Access to/from all Zones Coverage 100% Transit Fare Parking Costs - No Transfer Time - Other Usual Input Data Mode Choice Model Create Total Daily Transit Person Trip Table Create Total Daily Vehicle Trip Table All-or-Nothing Highway Assignment Capacity Unconstrained using VDF = (Link Distance in miles)/spd*60). Highway Assignment (Capacity Constrained) Summarize Results Constrained Highway Travel Times (MF cht ) Notes: 1. Average transit speed (SPD) was assumed to be 18 mph 2. A weight of 1.0 was used for transit walk time and wait time 3. See notes in Unconstrained Highway Model for performing Unconstrained Highway Assignment

19 Davuluri, Longo, Adury, Scott 17 Figure 9 Daily Transit Volume Comparison (No-Action Vs Unconstrained Transit)

20 Davuluri, Longo, Adury, Scott 18 Congestion Pricing Scenario Pricing strategies for congestion pricing can vary depending on the operating objectives. Since the primary goal of the current project was to find ways to alleviate congestion on a system wide basis, congestion pricing in this case was designed to focus on minimizing the overall network travel times and hence maximizing the system wide benefits. In order to maximize the system wide benefits all the facilities in the system need to be tolled and hence coming up with an appropriate toll pricing strategy for all the facilities in the system was an arduous task. For this study, system wide pricing was introduced using augmented Volume Delay Functions (VDFs) concept, which was developed based on the work of Dr. Randall Pozdena. The primary idea behind the augmented VDF approach is the application of marginal social cost in the system as opposed to the marginal individual cost that is usually perceived by the individual user. In a regular highway system the primary impedance perceived by individual users is their own delay; they do not perceive the incremental delay that their vehicles impose on other users. The key principle in the augmented VDF approach is that the toll paid by an individual user of the system should equal the incremental delay (cost) he/she is causing to the system. Hence, the total impedance (cost) to individual user is the sum of individual delay and the incremental delay. This methodology is implemented in the model by modifying the regular VDFs to the augmented VDFs. A regular VDF in the PSRC model is based on the Standard BPR (Bureau of Public Roads) formulation and it represents the individual delay component of the impedance. This VDF is modified to an augmented VDF and has two components, the individual delay component and the toll component. Hence, the toll in this case is applied in terms of time. A standard VDF function used in the PSRC model is of the type I own = t o [1+β(v/c) α ] Where, I own Impedance perceived by an individual user t o v/c Free flow travel time on a link Volume to Capacity Ratio α, β Constants that vary based on the facility type Based on the discussion in the above paragraph this VDF is modified to an augmented VDF as I total = I own + I toll = t o [1+(α+1)β(v/c) α ] Where,

21 Davuluri, Longo, Adury, Scott 19 I own = t o [1+β(v/c) α ] Impedance perceived by an individual user I toll = t o αβ(v/c) α ] Impedance due to pricing (incremental delay) Since the augmented VDF is a function of the V/C ratio for each link, appropriate pricing levels are automatically introduced on a facility as the volumes on that facility change. The PSRC model uses different α and β factors based on the facility type. Hence, varying levels of tolls are implemented based on the facility type. Under the augmented VDFs, greater impedance is perceived on the links and this causes changes in travel behavior, including alternative destinations with a shorter travel distance, diversion to alternative routes, and a shift to alternative modes. The reduction of travel on the most congested routes will improve their flow and operations, with the net effect being that of a system optimum condition for the auto mode. An important point to note that within this application of the regional model is that the overall person trip productions and attractions do not change in response to tolls; rather, the model redistributes demand in a different manner among alternative routes and modes, which will result in some person trips being shortened. Figure 10 shows the difference between the regular VDF and the augmented VDF for a PSRC model freeway link. Figure 10 Volume-Delay Functions for Private Own and Social Total Vehicle Marginal Delay Costs PSRC Model Freeway Link 7.0 Time (minutes / mile) Marginal Private Delay Marginal Social Delay The difference between the two curves represents the incremental delay factor = optimal toll time cost Volume-to-Capacity Ratio (V/C)

22 Davuluri, Longo, Adury, Scott 20 Methodology for Congestion Pricing Model Some of the basic assumptions that have been made in modeling this scenario are:- Fixed level of travel demand i.e. total productions and attractions of person trips remains constant Tolling is applied in terms of time Tolling is applied to only General Purpose vehicles. Transit Vehicles and HOV s are not be tolled. A modified/augmented volume delay function by facility type, as described previously, are be used to simulate tolls The augmented VDFs are not applied to HOV links. HOV links have regular VDFs. Tolls are applied on all auto mode links in the model. However, different classifications of facilities may have different toll values corresponding to v/c ratios and the VDF. Tolls are applied all day A regular model run with feedback loops was performed on the No-Action network using augmented volume-delay functions in the highway assignment process. During the first iteration constrained Non-HOV, HOV and transit travel times from No-Action were used as input into the Trip Distribution and Mode Choice steps. However, starting from the second iteration, special procedures were required to calculate the travel time to be input into Trip Distribution and Mode Choice. In a regular model run like the No-Action, the travel time outputs from the highway and transit assignments are directly used as inputs into Trip distribution and Mode Choice during the feed back loop. (Refer to Figure 3) This process cannot be exactly replicated in the pricing model as the travel times calculated from the assignments for the HOV and Transit modes would be inaccurate. Special techniques were required for the calculation of travel times for HOVs and Transit to be used as input into Trip Distribution and Mode Choice. By assumption HOVs and Transit vehicles were not tolled in this model. However, when HOVs and Transit Vehicles travel on links with the augmented VDFs, pricing is also applied to these modes. Hence, zonal and link level travel times calculated for these modes from the regular assignment would be inaccurate as the travel times reflect the toll in terms of time. To account for this phenomenon, pure (toll free) travel times were calculated for HOVs and Transit based on special assignments that used the regular (toll free) volume delay functions. (The procedure for the calculation of pure travel times has been described following this paragraph.) The above calculated pure times for HOV and Transit modes along with the augmented travel times for Non-

23 Davuluri, Longo, Adury, Scott 21 HOVs (output from a regular highway assignment) were used as input into Trip Distribution and Mode choice. This procedure was repeated for desired number of iterations. At the completion of the feedback loop process, pure times were calculated one last time for all the modes. These pure times were used to summarize the results and to calculate various performance measures for this model run. The flow chart shown in Figure 11 highlights the methodology used for the pricing model. Production of pure travel times For the purpose of calculating the pure travel times no modifications to trip tables or assignment results produced from a model run with augmented VDFs was necessary. An additional options assignment was performed to calculate the pure travel times. (1) The volumes obtained from the assignments using augmented VDFs are saved into a link extra attribute such as UL3. (2) The augmented VDFs were replaced with the toll free VDFs. However the volau part of the regular VDF was replaced with volad. (3) A zero iteration additional options auto assignment was performed with the link attribute containing the original volumes (UL3) as the source for additional auto volumes on the links. Link and zone-to-zone pure highway travel times are produced from this step. (4) A standard transit assignment was then performed on this network that has toll-free times on all the links. This transit assignment utilized the pure congested auto times to calculate pure transit travel times. Application of Enif for the Congestion Pricing model run Several plots produced using Enif were used to understand and analyze the impact of pricing in the region. Figure 12 shows the daily volume differences between the Pricing Model and the No- Action. This plot shows the effect of pricing in the system. Volumes were reduced on all the major freeways in the system, while volumes increased on the arterials. This resulted in a better utilization of highway capacity. On a system wide level it was found that pricing was beneficial in reducing the over all system delay. Several performance indicator plots (similar to the ones produced for No-Action) were produced to evaluate the performance of various facilities under pricing. These plots were used as guideline to determine the corridors that need to be tolled in some of the study alternatives.

24 Davuluri, Longo, Adury, Scott 22 Figure 11 Modeling Methodology for Congestion Pricing Model SCHEMATIC FLOW CHART 2025 CONGESTION PRICING SCENARIO Using 2025 No-Action Scenario with Augmented Volume Delay Functions Latest 2025 Land Use Forecasts Trip Generation Model: -Trip Ends by Trip Purpose (P s & A s) Trip Distribution Model (Constrained) Parking Costs Transit Fare - Other Usual Input Data Mode Choice Model Highway Assignment (Capacity Constrained) Travel Times for Non-HOVs Reflecting Augmented VDFs Transit Assignment Using Toll Free VDFs, Calculate Pure Travel Times for HOVs & Transit Calculate Pure Travel Times for Non-HOVs, HOVs & Transit using Toll Free VDFs Summarize Results Note: Apply modified volume-delay functions (VDFs) to all links excepting to HOV links, zone connectors & ferry links.

25 Davuluri, Longo, Adury, Scott 23 Figure 12 Difference in Daily Vehicle Volumes (Pricing Scenario minus No-Action)

26 Davuluri, Longo, Adury, Scott 24 CONCLUSION This paper described an application of EMME/2 and Enif for a Congestion Relief Analysis Study in the Puget Sound Region. The study looked at various multi-modal alternatives to mitigate congestion in the region. These study alternatives were determined based on the results obtained from unconstrained highway, unconstrained transit and the pricing scenarios. EMME/2 related special modeling techniques used to model these three scenarios have been discussed in the paper. The paper also described the use of Enif as an analysis/evaluation tool in analyzing the results from these model runs and in defining the study alternatives. Enif has greatly enhanced the visual representation of the data contained in EMME/2. It allowed a better utilization of the information stored in the EMME/2 databanks. Though the usage of ENIF for this study was not extensive, it turned out to be highly beneficial to the extent it was used. The CRA study is currently ongoing; several analyses are being carried out and hopefully, results from these analyses will be presented at the next EMME/2 Users Conference. REFERENCES 1) Puget Sound Regional Council Transportation Pricing Alternatives Study - Technical Memorandum 3, Simulating Congestion Pricing in EMME/2 by Dr. Randall Pozdena, ECONorthwest, February 19, ) Extending Highway A-25: An Exploratory Analysis with EMME/2 and Enif - Isabelle Constantin, INRO Consultants Inc., and P. Maillard, MTQ, Montreal, Quebec, Canada, Presented at the 16th International EMME/2 Users' Conference, March ) Urban Areas Congestion Relief Analysis - No Action and Unconstrained Modeling Methodologies Working Paper February 4, ) Urban Areas Congestion Relief Analysis - Bookends Modeling Methodologies Working Paper March 29, ) Urban Areas Congestion Relief Analysis Central Puget Sound Draft Technical Report October 7, ) EMME/2 Users Manual, Release 9 7) Getting Started with Enif: An Introductory Guide to Enif, Based on Version 1.5 ACKNOWLEDGMENTS The authors would like to thank Youssef Dehghani, Mark Scheibe, and Cathy Strombom of Parsons Brinckerhoff Quade & Douglas, Inc. as well as Don Samdahl of Mirai Associates for their assistance and guidance during the course of the Congestion Relief Analysis (CRA) Study. We also would like to thank Shuming Yan of Washington State Department of Transportation (WSDOT) for his guidance and support. Opinions and views presented in this paper are solely ours and do not necessarily represent the views or official policy of the WSDOT and its staff.