Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

40 C H A P T E R 4 4.1 Introduction Presented herein are three case studies that illustrate various capacity assessment methodolo- gies to determine if adequate line capacity exists given specific assumptions about the train mix, volume, and operating patterns. The first case study, a conceptual grid time analysis of the LOSSAN Corridor in South- ern California, served as a means to help public agencies along the corridor understand the range of capacity enhancements required over time to ensure fluid passenger and freight operations. The second study relates the results of an RTC operations simulation for the start-up and build-out of the New Haven-Hartford-Springfield (NHHS) commuter rail service. That service will share the Amtrak Springfield Line with Amtrak and freight train traffic starting in 2016. The third case study compares findings of an RTC simulation, a grid time analysis and an application of the NCFRP web-based SU Tool brought to bear on a start-up of proposed regional passenger rail operations along the North Puget Sound (hereafter, North Sound) between Bellingham and Everett in Washington State. As study team members for this guidebook had worked on the LOSSAN, NHHS, and North Sound analyses, these cases were selected to illustrate the various capacity assessment methodologies. 4.2 LOSSAN Corridor Capacity Investment Planning 4.2.1 Introduction The 351-mile-long LOSSAN Rail Corridor between San Luis Obispo, Santa Barbara, Los Angeles and San Diego is the second busiest passenger rail corridor in the U.S., second only to the Boston-to-Washington Northeast Corridor. More than 7.2 million passenger riders make trips on LOSSAN Corridor trains annually. (Reference: LOSSAN Corridor Strategic Implementa- tion Plan, San Diego Association of Governments, April 2012.) Looking toward a future of higher gasoline prices and more congestion on parallel road systems, the demand for the corridor’s rail service is likely to grow. The 2010 LOSSAN Corridor Strategic Assessment, sponsored by the Orange County Transpor- tation Authority and Caltrans, was in part an attempt to identify the rail line capacity constraints and the scope of potential solutions needed to maintain adequate capacity for passenger and freight trains in the corridor for the foreseeable future. The methodology utilized was a concep- tual grid time analysis. Best Practices

Best Practices 41 4.2.2 Existing Corridor Services There are four different corridor passenger rail services. These are: • The Pacific Surfliner, operated by Amtrak with financial support from Caltrans Division of Rail, between San Diego and San Luis Obispo via Los Angeles Union Station (LAUS). • The Metrolink commuter rail service, operated by the Southern California Regional Rail Authority (SCRRA) between Oceanside and Montalvo (north of Oxnard) via LAUS. • COASTER commuter rail, operated by North County Transit District (NCTD) between San Diego and Oceanside. • Amtrak Long-Distance Network Services: the Coast Starlight operating between Seattle, northern California and Los Angeles; and the Southwest Chief operating between Chicago and Los Angeles (for a relatively short segment between Fullerton and LAUS). There are three freight rail operators on the LOSSAN Corridor, sharing track with passen- ger trains. UP serves customers between San Luis Obispo and Los Angeles, and between South Anaheim and Santa Ana. BNSF runs trains between Los Angeles, Fullerton, and San Diego. A short line or small railroad, the Pacific Sun Railroad, serves local customers in the Oceanside area. On a typical weekday, there are as many as 100 trains per day on the busiest portion of the corridor, between Redondo Junction near Downtown Los Angeles and Fullerton. 4.2.3 Planning for the Future The growth of corridor ridership has been dramatic. In 1979, the Amtrak San Diegans carried 1.2 million passengers. Ten years later, ridership totaled 1.8 million. Metrolink commuter rail service started in 1992, followed by COASTER in 1995. In 2000, the San Diegans were renamed Pacific Surfliners to more accurately capture the range of its service, which by then extended to San Luis Obispo. All three services have expanded to meet the ever growing demand, which is now six times what it was 30 years ago. Continued ridership and service growth, however, face challenges. Chiefly among these is that higher numbers of trains are reaching the capacity limits of the physical plant. There have been many studies of the LOSSAN Corridor and its capacity needs. The original planning work began in the 1980s. In the time since, passenger rail operators separately have developed service expansion plans, but these studies have been service-specific. The LOSSAN Rail Corridor Agency, the Joint Powers Authority charged with coordinating planning efforts for the corridor, identified various long term investment options to support more passenger rail service. However, the improvements were not tied to specific increases in train traffic over time. The LOSSAN Corridor Strategic Assessment aimed to make that link of improvements to train volumes. The first step was to assess the state of the corridor. Current passenger and freight operations were profiled. Second, funded or programmed capital investments in the corridor’s physical plant were identified. Lastly, operating conditions on the corridor were assessed, with capacity bottlenecks identified. Figure 4-1 shows corridor weekday train volumes at the time of study initiation in 2008. 4.2.4 Grid Time Analysis The study required a basic understanding of where line capacity problems exist and where more trains might be added, given both existing conditions and planned or programmed, near term line capacity improvements. The tool to enable this understanding was the capacity “yard

42 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations Figure 4-1. Line densities on LOSSAN corridor segments in 2008. Track Configuraon Minutes between Sidings Minutes of Headway Praccal Capacity in Trains per Day Single Track 15 50 20 35 30 25 Double Track 10 150 20 75 Table 4-1. Practical rail line capacities. sticks” appearing in Table 4-1. The methodology employed relates to the grid time analysis described in Chapter 3. The practical capacity limits of two track configurations are calculated Table 4-1. Here, practi- cal capacity is defined as the number of trains that can run on a track segment efficiently given its configuration and appropriate allowances for both regular maintenance-of-way and small, incidental occurrences that work to delay trains. In other words, practical capacity is the point at which the addition of new trains begins to degrade operating performance on a specific corridor segment. For example, the practical capacity of a single track segment with frequent sidings is calculated as follows: • Seven or eight miles between sidings equates to about 15 minutes of one-way grid time between sidings, given an average freight speed of 30 miles per hour.

Best Practices 43 • With 15 minutes between sidings, maximum capacity equals four trains per hour between sidings. • Maximum daily capacity or theoretical capacity equals four trains per hour multiplied by 24 hours or 96 trains per day. • Practical daily capacity would be half that figure, or about 50 trains per day. Conceptually, any trains above this number could negatively impact performance. This formula provides for maintenance-of-way, random delaying occurrences, and the mix of train types that traverse the corridor. In this context, the halving of theoretical capacity to bracket practical capacity is based on practical experience with schedule and train performance variability for the kinds of traffic on a line. It includes assumptions of long, heavy and slow freight trains that do not operate on sched- ules, along with fast and light schedule passenger trains. With a highly disciplined operation, practical capacity can begin to approach theoretical capacity. An example is a big city transit operation versus a Class I medium density line. The former gets more out of the capacity that is available. The maximum capacity of a double track segment is calculated differently: • Assumed braking distance for large freight trains (operating on all corridor segments) is two miles. • Assuming a simplified signal system, another two miles is required to stop a train. • Assuming 25–30 mph freight train speeds and a four-mile braking distance, minimum head- ways would be around eight to 10 minutes between trains. • With 10 minutes between trains, maximum capacity equals six trains per hour. • Maximum daily capacity or theoretical capacity equals six trains per hour multiplied by 24 hours, or 144 trains per day per track or 288 per double track. • Practical daily capacity would be half that figure, or about 150 trains per day, sufficient to allow for maintenance-of-way, random delaying occurrences, and traffic mix. • It should be noted, that except for the short segment between Fullerton and Redondo Junc- tion, south of LAUS, the predominant use of the corridor is passenger trains, with relatively similar operating characteristics and maximum speeds, but with widely differing stopping patterns (intercity versus commuter). The advantage of such a conceptual approach in estimating practical line capacity is that it is straightforward and fairly simple to do. One has to know train counts and track configurations. But the conclusions on capacity rest on a number of assumptions about all train movements on a specific track segment. This is not always the case. Train type, speed, and length vary and, as a result, a specific segment of track may have more or less of a practical capacity limit than the table above indicates. Nevertheless, the approach helps to point out where opportunities and trouble spots might occur. The advantage of operations simulation versus a conceptual grid time analysis like the LOSSAN study is it can deal easily with a multitude of variations. It is most usefully employed when projects are closer to being realized. The reason is, operations simulation is time consuming and expensive to undertake. This is to say, the dollars are better spent when the desired outcomes are better defined. 4.2.5 Future Train Volumes and Required Improvements Table 4-2 compares the practical capacity of LOSSAN Corridor segments with the estimated future traffic volumes. In four cases, the estimated 2020–25 train volumes will be below the practical capacity for the line segments. However, for the remainder of segments, future volumes

44 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations will be greater than the estimated practical capacities of the segments, indicating that capac- ity enhancements for these line segments will be needed at some point. Train volumes for the intervening years were calculated (these were not shown in the table for the sake of simplic- ity) so implementation of improvements could be identified in five-year increments (2010–15, 2015–2020, and 2020–25). A few examples are listed herein from north to south. • Santa Barbara-Ventura: Sidings improvement will be required by 2011–15. Practical capacity today is about 25 trains per day. Specific improvements could include Seacliff Siding north and Rincon Siding. With the implementation of Santa Barbara commuter service in the near term, a north platform would be needed at Oxnard by 2011–15 as well. • Ventura-Van Nuys: By 2020–25, more sidings and/or double track will be required to ensure capacity for at least 72 trains per day. • Los Angeles-Fullerton: Triple track as soon as possible. While the segment may not have reached its practical capacity limit, there is little to no room today for more peak period service. Qua- druple track will be required by 2020–25. • Laguna Niguel-Oceanside: Siding improvements and some double tracking will be required by 2020–25. • Oceanside-San Diego: Some siding improvements will be required by 2011–15. More sidings, double track and tunnels will be required by 2020–25. 4.2.6 LOSSAN Grid Time Analysis Summary The LOSSAN study fell short of recommending specific locations and specific types of improvements. That sort of specificity would be the product of a more detailed examination, using rail operations simulation, when actual future train volumes and schedules are better defined. With this analysis, the study aimed to illustrate for LOSSAN policy makers where capac- ity constraints will likely exist given certain assumptions of daily train volumes and the types of solutions that could be deployed. The 2010 study addressed more issues than infrastructure needs. These included an endorse- ment of a hierarchy of services, ranging from express and limited stop intercity services, and inter- regional commuter services involving equipment belonging to both Metrolink and COASTER, improved train connectivity at stations, and common fare instruments. These improvements were seen as means to make the corridor more convenient to use and thus spur ridership. How- ever, it was the linking of infrastructure improvements to train volumes over time that was at the heart of the LOSSAN strategic vision—an outcome realized through the use of the conceptual grid time analysis described above. LOSSAN Segment Praccal Capacity 2008 Volume (Baseline) 2020 2025 Volume San Luis Obispo – Santa Barbara 35 14 20 Santa Barbara – Ventura 25 20 34 Ventura – Moorpark 35 28 72 Moorpark – Van Nuys 50 42 72 Van Nuys – Burbank Juncon 150 44 74 Burbank Juncon – Los Angeles 150 85 134 Los Angeles – Fullerton 150 102 232 Fullerton – Orange 150 45 104 Orange – Laguna Niguel 150 65 152 Laguna Niguel – Oceanside 50 44 56 Oceanside – San Diego 50 48 98 Table 4-2. Practical capacity versus baseline and future volumes.

Best Practices 45 Over the intervening years, conditions on the corridor have changed in numerous ways. Freight train growth, for one thing, has been slower than anticipated, an outcome of the recent economic recession and curbing of rail-borne international container volume going to and from the Ports of Los Angeles and Long Beach. Commuter train operations and growth assumptions have changed as well. Since the grid time analysis, the LOSSAN agency has continued to study the timing and location of improvements on the corridor through operations simulation using RTC with current assumptions for passenger and freight train operations. 4.3 New Haven-Hartford-Springfield Corridor Planning 4.3.1 Introduction Since 2001, the Connecticut Department of Transportation has been working toward the implementation of a new commuter rail service on the 61-mile Amtrak Springfield Line between New Haven, Hartford and Springfield, MA. Preliminary work on what was to be called the New Haven-Hartford-Springfield (NHHS) Corridor was completed in the first half of the decade. Then in 2008, work began on an Environmental Analysis, which was completed and accepted by the Federal Railroad Administration in 2012. A map of the line appears as Figure 4-2. Figure 4-2. New Haven-Hartford-Springfield (NHHS) line.

46 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations A key element of all work phases since the beginning of the project was the use of rail opera- tions simulation to identify line capacity enhancements required to ensure fluid passenger and freight rail operations on the line, which is also a federally designated high speed rail corridor. The simulation task was performed by means of the Rail Traffic Controller software, which is a standard tool for rail operations analysis and train performance evaluation. (Reference: 2016 Start-Up HSIPR Corridor Service, Version 2d, produced by CDM Smith, 2012.) The proposed New Haven-Hartford-Springfield High Speed Rail Corridor is planned to be implemented in phases. This phased implementation approach will require adaptation of operating plans and schedules to utilize as much of the added capacity as possible to provide improved passenger rail services without reducing the performance of the freight railroads. 4.3.2 Existing Corridor Operations Existing train operating information was provided by the operators on the route. These were: Amtrak Northeast Corridor passenger trains, Pan Am Railways (PANAM), CSX Transportation, Providence and Worchester Railroad (P&W), and the Connecticut Southern Railroad (CSO) freight trains. Conversations with all carriers occurred in the late summer and early fall of 2008 and an update, performed in mid-2011, identified an overall decrease in scheduled freight service. The existing passenger schedules were based on Amtrak’s summer/fall 2011 schedule and included the Amtrak trains operating on the Springfield Line, and Amtrak Northeast Corri- dor (NEC) trains and ConnDOT sponsored Shore Line East (SLE) commuter trains operating between New Haven and Mill River, the NEC junction for the Springfield Line. 4.3.3 Simulation Parameters 4.3.3.1 Modeled Operating Cases A comparison of the modeled cases is shown in Table 4-3. Each of the listed cases was coded into the RTC simulation software, and a set of ten weekly schedules was simulated. Random Simulaon Case Descripon Infrastructure Improvements Schedule Train Ranking/Priority Case 1 No Build 2011 Exisng 2011 Amtrak 2011 Freight Higher Priority Passenger Trains Case 2 No Build 2030 Exisng 2011 Amtrak 2011 Freight Grown to 2030 Levels at 1.75% per Year. Higher Priority Passenger Trains Case 3 2016 V2d (2030 Build) Improved Double Track between Cedar Hill Yard and Har‹ord Staon and between Har‹ord Yard and Hayden Interlocking, Addional Siding and Rehabilitaon Running Track in Har‹ord Yard, Upgrade Switches Har‹ord Yard, Hayden Interlocking. Amtrak 2016 V2d Service Plan Including Expanded Intercity and New Commuter Service. 2011 Freight Grown to 2030 Levels at 1.75% per Year. Higher Priority Passenger Trains Table 4-3. Modeled operating cases.

Best Practices 47 variations to scheduled dwell times and initial departure times were applied to simulate the effect of minor random day-to-day impacts on train operations. The results were averaged over the runs and are presented below. Simulation results were used to evaluate the performance of the pro- posed service plan in combination with the improvement of the existing rail infrastructure, compared to the existing infrastructure and service plan. 4.3.3.2 Train Priority and Ranking One of the key features of RTC is its meet-pass conflict resolution logic. RTC resolves “meets” or conflicts of opposing trains on the basis of priority, just as a human dispatcher would. For example, if a passenger train has a higher priority than a freight train, then when a passenger train and a freight train are approaching each other on single track, the passenger train would “hold” (remain on) the main line while the freight train would “take” (enter) a siding in order to let the passenger train pass. Also, if a train is running late, its priority increases. The opposite is true if a train is running early on its schedule. There are three priority values to be set in RTC: minimum, initial, and maximum. The initial priority is a value assigned to a train when it goes on line; minimum and maximum are the lowest and highest boundaries. RTC also offers a second layer of dispatching criteria based on train ranks. This parameter is used to handle special trains such as high and higher speed rail. There are seven ranks available: ranks 1 to 3 designate train types as elite, while 4 to 7 as regular. RTC will strive to keep higher ranked trains on schedule; lower ranked trains can be forced to take large delay in order to keep elite trains on schedule. 4.3.3.3 Randomization All simulations were modeled to recognize that there is a level of randomness that occurs in train operations. The RTC software can recognize this with the application of randomization fac- tors applied to each case. For the simulations the initial departures and dwell times were allowed to vary on a random basis within the following parameters: passenger trains were modeled with up to 2 minutes late initial departure and up to 15 seconds extended dwell time. Freight trains were modeled with up to 15 minutes early/late initial departure and up to 5 minutes extended dwell time. 4.3.3.4 Simulation Run Settings Each simulation was run for a 7-day period plus half a day for warm-up and cool-down. Ten runs were performed for each case, and each run had a different random seed, viz., train sched- ules and delays were different during each run. The results of the 10 runs were averaged and reported as the results for each particular operating case modeled. 4.3.3.5 Train Performance Calculator The train performance calculator (TPC) parameter depicts the ideal run of a train. TPC run times assume no conflicts with other trains, all switches aligned, and all green signals along the route. RTC trains were calibrated to replicate performance reported on TPC charts provided by Amtrak. Also speed restrictions on the corridor were coded according to Amtrak’s train perfor- mance chart, which showed a maximum design speed of 110 mph with 80 mph speed limits at level crossings. The speed restrictions north of Hartford Station were assumed to remain identi- cal to the current speed limits. 4.3.3.6 Pad According to FRA’s Railroad Corridor Transportation Plans guidance manual, whenever pas- senger schedules are produced by various TPC runs, a pad (make-up/recovery time) must be

48 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations added to the TPC schedule to account for a number of factors. For a double track network the schedule pad is calculated by increasing the train runtime by a minimum of 7% to take into account such factors as human operation instead of perfect TPC operation, temporary slow orders, congestion or off-schedule trains, adverse weather conditions, and signal imposed delay, etc. 4.3.3.7 On-Time Performance The On-Time Performance (OTP) parameter represents train schedule adherence and is expressed as the percentage of trains arriving between their scheduled arrival time and a specific OTP threshold value. OTP threshold values used in the simulations are shown in Table 4-4. For passenger trains values are based on the minimum pad (7%) for the longest passenger running time, specifically, six minutes pad for a scheduled run time of one hour and 30 minutes. With respect to freight trains, considering their length/weight and conse- quent slower acceleration/deceleration capability, 15% of the longest service runtime train was used. Exceptions are NEC and SLE passenger trains, which traverse the network for approximately 15 minutes, thus the lower threshold. The same reasoning was applied to freight trains with short running time such as the CSXT trains between West Springfield Yard and Springfield Station. 4.3.3.8 Train Frequencies Freight service frequencies and schedules were identical to the existing 2011 schedules for all cases. For the simulations in 2030, it was assumed that freight train lengths and weights will increase by 1.75% annually over the next 19 years (a 39% increase over 2011) for both the No Build and the 2016 V2d cases. The service frequencies are shown in Table 4-5. Passenger Services Train Type Train Runme in Simulaon Corridor (Minutes) OTP Threshold (Minutes) New Haven – Springfield 90 6 NEC and SLE 15 2 Freight Services Train Type Train Runme in Simulaon Corridor (Minutes) OTP Threshold (Minutes) CSO 420 60 CSXT Springfield 30 5 Table 4-4. On-time performance (OTP) threshold values. Case Passenger CSO CSXT PANAM P&W All Trains NHV HFD SPG 2011 No Build 92* 18 15 2 18 145 2030 No Build 92* 18 15 2 18 145 2016 V2d 245* 18 15 2 18 298 *Note: includes passenger trains running the full 61 mile corridor and those only traversing a small sub segment near New Haven staon. Table 4-5. Comparison of frequency of train service in trains per week.

Best Practices 49 4.3.4 Simulation Results Due to the increase in freight train length and weight, the performance parameters do show a slight degradation of operating performance when evaluating 2030 No Build versus the 2011 No Build Case. The delay statistics for 2016 V2d (2030 Build) simulations are unchanged or improve relative to the No Build cases except for the Delay Percentage for freight. On-time performance improves relative to the 2030 No Build case for both freight and passenger, even though volumes of the line more than double. Overall delay and on-time metrics improve. Table 4-6 shows a comparison of performance parameters. In addition, Figure 4-3 shows the location and the amount of delays occurring in the network. The 2030 Build case clearly shows significant reduction in the delays in the improved (double track) section between the Mill River junction and Hartford, compared to the No Build cases. Figure 4-4 shows the cumulative percent delay for both passenger and freight trains. For exam- ple, 95% of passenger trains have cumulative true delay (delay in run time) of less than 5 minutes. It can be observed that the 2030 Build case passenger and freight cumulative delays are compa- rable to the 2011 No Build case, without any optimization to the 2030 Build case train schedules. 4.3.5 NHHS Operations Simulation Summary To conclude, the increased passenger service, as proposed in the 2016 V2d corridor service in combination with the assumed growth in freight train length and weight due to future demand in rail shipments, can be handled on the proposed infrastructure with sufficient operational Delay Percentage (1) (4) Case Passenger Freight Overall Delay Percentage (1) (4) 2011 No Build 1.4 18.1 7.2 2030 No Build 1.6 18.4 7.7 2030 Build (2016 V2d) 1.3 18.8 5.0 Minutes of Delay per 100 Train Miles (2) 2011 No Build 1.7 54.1 11.3 2030 No Build 2.0 58.2 12.3 2030 Build (2016 V2d) 1.7 53.3 6.2 On-Time Performance (3) 2011 No Build 97.9% 100.00% 98.7% 2030 No Build 97.2% 99.2% 97.9% 2030 Build (2016 V2d) 99.2% 99.6% 98.9% 1. Delay percentage is measured as the rao between the amount of delay at the terminal staon versus the total amount of scheduled train run me 2. Rao of total amount of delay of all trains and the amount of train miles traveled in the simulaon corridor. 3. Trains are assumed to perform on me when the arrival me at the terminal is less than 6 minutes later than the schedule arrival me for passenger trains and less than 1 hour for freight trains. 4. Delay is defined as the total difference between the planned arrival me and the recorded arrival me at the terminal staon. The results are averages of 10 weekly schedules simulated using randomizaon. Table 4-6. Comparison of performance parameters.

50 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations quality. Optimization of the freight service can potentially provide further reductions in delay for both passenger and freight service. 4.4 North Sound Rail Assessment Comparisons 4.4.1 Introduction In 2010 North Sound Regional Rail Operations Simulation, sponsored by the Whatcom Council of Governments, was initiated to answer the following question: Are the capital improvements previously identified for the implementation of new Cascades service sufficient to enable the imple- mentation of reliable commute-oriented regional rail passenger services between Bellingham and Everett, WA, on the BNSF Railway? Figure 4-3. Operating delay analysis.

Best Practices 51 This question was first explored in a 2008 analysis. The 2010 study was an update. Both studies followed the same approach: use RTC to test if existing and proposed anticipated pas- senger plus existing freight trains can share the line without serious deterioration in service quality. This case study first details the results of that operations simulation answering the question posed above. Subsequently, the same basic question is investigated using two other methodologies—a grid time analysis and an analysis employing the NCFRP web-based operations simulation tool—and the results of all three are compared. The case study concludes with observations of using three different tools for the same job. 4.4.2 North Sound Rail Operations Simulation 4.4.2.1 Introduction The work scope for the 2008 and subsequent 2010 studies included a computer-based simula- tion of railroad operations with and without the proposed regional rail trains. The study area for the 2010 study was the same as for the 2008 study: Wenatchee to Everett, Seattle to Everett, and Everett to Bellingham and Blaine as shown in Figure 4-5. BNSF provided records of actual train movements for a month-long period that were used by the study team to create train files representative of current traffic patterns, including time and day of operation, train length and tonnage, and crew change locations. (Train files are used to Figure 4-4. Cumulative percent true delay—passenger and freight trains. 75% 80% 85% 90% 95% 100% 0 5 10 15 20 25 30 Cu m ul a ve Pe rc en ta ge of Tr ai ns True Delay (Minutes) Cumulave Percent True Delay—Passenger Trains 2011 No Build 2030 No Build 2030 Build 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 100% 0 10 20 30 40 50 60 Cu m ul a ve Pe rc en ta ge of Tr ai ns True Delay (Minutes) Cumulave Percent True Delay—Freight Trains 2011 No Build 2030 No Build 2030 Build

52 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations perform the simulation.) The 2010 study took into account the then current Sounder, Amtrak Empire Builder and Cascade trains, plus a third Cascade round trip between Seattle and Vancouver, British Columbia. This train was assigned a mid-day schedule as envisioned by the Washington State Department of Transportation. The Rail Traffic Controller program was used for this analysis. RTC is also used as an analysis tool by BNSF. All simulations were performed with RTC Version 2.70 L59E. The regional rail service plan assumed in the simulation would have two morning trains departing Bellingham and continuing to Everett, with intermediate stops at Mt. Vernon-Burlington, Stanwood, English, and Marysville. At Everett, the trains would connect with Sounder com- muter service between Everett, Mukilteo, Edmonds, and Seattle King Street. Two corresponding afternoon trains would leave Everett after arrival of connecting Sounder trains, and return to Bellingham. In order to include the collateral effects of freight service outside the immediate Bellingham-Everett service area, the simulation area included trackage of the Bellingham Sub- division from the Canadian border at Blaine south to Everett, and trackage of the Scenic Subdivi- sion from Wenatchee west to Everett and Seattle. The regional rail schedules assumed appear in Table 4-7. Figure 4-5. North sound regional rail operations simulation study area.

Best Practices 53 4.4.2.2 Network Simulations The rail network for the simulations was drawn using current BNSF track charts and time tables (variously dated between 2007 and 2010), provided by BNSF simulation modelers. Cen- tralized Traffic Control signalization (whereby a dispatcher in a remote location directs the prog- ress of trains over a section of track by wayside signals) was added to the simulation cases except in yard areas. Assumptions regarding the track and capacity improvements for each simulation case were taken from the 2008 study; these were based on diagrams of conceptual improvement plans prepared for WSDOT by Transit Safety Management and HDR Engineering. The improvements represent a package of track improvements that WSDOT and BNSF had accepted as require- ments for extension of the second Amtrak round trip to Vancouver, the addition of a third Amtrak round trip, as well as for increasing Sounder service to four round trips between Seattle and Everett. Some of the improvements assumed in the 2008 study had since been realized. Others are scheduled to be completed with American Recovery and Reinvestment Act (ARRA, 2009) funding or other funds in the near future. Thus, assumed for the base case (Simulation 1) were: • Double track through Interbay Yard • Double track from Milepost (MP) 7.3 to MP 7.8, with universal crossover at MP 9.0 • Incorporation of Mukilteo Sounder station • Extension of Lowell siding west to MP 1783.0 • Extension of English siding southerly to MP 43.9 • Extension of Stanwood northerly to MP 57.6 • Extension of Mt. Vernon siding southerly to MP 65.5 • Burlington yard revisions and new main line (no siding at this location) • Revision of Custer and Intalco sidings to provide extended yard track (no siding at this location) • Extension of Swift siding between MP 114.9 and MP 118.1 with 5 mph freight speed limit at north end NORTHBOUND – PM SOUTBOUND AM Staon NS 01 NS 03 Staon NS 02 NS 04 Evere 17:45 18:45 Bellingham 5:10 5:40 Marysville 18:06 19:06 Mt. Vernon 5:41 6:11 English 18:12 19:12 Stanwood 5:56 6:26 Stanwood 18:22 19:22 English 6:06 6:36 Mt. Vernon 18:37 19:37 Marysville 6:12 6:42 Bellingham 19:05 20:05 Evere 6:30 7:00 Note: (1) Intermediate dwells 30 seconds. (2) Departure „mes shown for all sta„ons, with excep„on of terminals (arrival „me). Table 4-7. Regional rail schedules.

54 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations Improvements in the vicinity of Delta Yard, identified and modeled during the previous study, were not included in the 2010 effort. Specifically: • Ease of main line curve at north end of the yard • New main line around the yard • Revision of yard tracks/leads However, by means of distance equations, study modelers coded longer receiving and depar- ture tracks at Delta Yard to avoid long freight trains blocking yard approaches during crew change layovers or switching operations. Train files, as noted, were based on BNSF records of actual movements during the peak week of October 2010. Based on train type, thresholds for early or late departure were included, and the RTC random feature was used to simulate different departure times for each day of opera- tion. RTC was also set to permit variations in dwell time. The simulations were run for a seven- day statistical period, with 24-hour warm-up and cool-down periods. These periods are a programming feature that is to assure steady state of operations for the simulation. While a simulation may start at a specific time, e.g. 8 AM on a Monday, in reality trains are moving on the line before the start time and dealing with conflicts from opposing trains. A warm-up period accounts for train operations before the start time of a simulation and thus helps to ensure a steady state of operations by the time the simulation actually begins. The function of a cool-down period is the same, only pertaining to the end of a simulation. Statistical performances were calculated by averaging ten simulation runs for each simula- tion case. The study team provided BNSF with draft final RTC simulation and summary files. After addressing all comments by BNSF staff, cases were re-run, obtaining the results summarized later in this section. Four cases were simulated: • Simulation 1 represents existing track configuration, plus near term track improvements, and existing trains, including the fourth Sounder round trip between Seattle and Everett, two Cascade round trips between Seattle and Vancouver, the Empire Builder, and BNSF trains. • Simulation 2 has the same track configuration and freight trains as Simulation 1. However, with respect to passenger trains, it adds the two Bellingham-Everett weekday regional rail trains and a third Cascade Seattle-Vancouver round trip on a mid-day schedule. • Simulation 3 adds a set (Set 1) of capacity improvements, which include double tracking two sections between Everett and Seattle (MP 27.0 to MP 27.8 and MP 15.8 to MP 17.8), and implementing two universal crossovers (at MP 27.8 and MP 17.8). • Simulation 4 adds a further set (Set 2) of track improvements by extending the siding at South Bellingham, joining Samish and Bow sidings, and implementing a universal crossover at MP 81.0. Table 4-8 shows the list of track improvements and trains operating between Everett and Bellingham. 4.4.2.3 Analysis of Simulated Performance Shown in Table 4-9, the statistical performance measures, by which the simulation cases can be compared, are defined herein. Average Train Speed: Average passenger train speed increases with the introduction of the new regional rail service; that is, due to more trains running on the Bellingham Subdivision, which has higher speed limits than most sections in the Scenic Subdivision (especially along the mountain- ous line between Everett and Wenatchee), speeds on average increase. Capacity improvements

Best Practices 55 Improvement Simulaon 1 Simulaon 2 Simulaon 3 Simulaon 4 Freight Service Current Service Current Service Current Service Current Service Cascade Service and Amtrak Empire Builder 2 Vancou’r RT 1 Em. Bldr. RT 3 Vancou’r RT 1 Em. Bldr. 3 Vancou’r RT 1 Em. Bldr. RT 3 Vancou’r RT 1 Em. Bldr. RT Commuter and Regional Rail Service 4 Sounder RT 4 Sounder RT 4 Sounder RT 4 Sounder RT 2 Regional RT 2 Regional RT 2 Regional RT Construct regional rail layover tracks at Bellingham (MP 94.4) and Everett (MP 1782.7) No Yes Yes Yes Double track, MP 27.0 to MP 27.8 with universal crossover at 27.8 No No Yes Yes Double track MP 15.8 to MP 17.8 with universal crossover at MP 17.8 No No Yes Yes Extend South Bellingham northerly to MP 97.0 No No No Yes Join Samish and Bow with universal crossover at MP 81.0 No No No Yes Table 4-8. Assumed track improvements for north sound regional rail simulations. Table 4-9. Simulated performance for 7-day period. Measure Simulaon 1 Simulaon 2 Simulaon 3 Simulaon 4 Base Case Base Case + New Service Base Case + New Service + Improvements Set 1 Base Case + New Service + Improvements Sets 1 and 2 Passenger Train Count 82 116 116 116 Expedited Train Count 72 72 72 72 Freight Train Count 240 240 240 240 Total Train Count 394 428 428 428 Passenger Train Miles 14,096 16,696 16,726 16,774 Expedited Train Miles 117,775 117,801 117,711 117,748 Freight Train Miles 195,890 195,793 195,327 195,699 Total Train Miles 327,761 330,290 329,764 330,222 Average Passenger Speed 35.6 mph 36.5 mph 36.6 mph 36.6 mph Average Expedited Speed 22.8 mph 22.7 mph 22.7 mph 22.8 mph Average Freight Speed 14.9 mph 14.6 mph 14.7 mph 14.9 mph Overall Average Speed 19.2 mph 19.7 mph 19.8 mph 19.9 mph Passenger Delay Percent 3.8% 4.7% 4.4% 4.2% Expedited Delay Percent 18.6% 18.9% 18.9% 18.2% Freight Delay Percent 33.6% 36.8% 35.6% 33.4% Overall Delay Percent 24.7% 25.6% 24.9% 23.6% Passenger Delay per 100 TM 5.4 minutes 6.4 minutes 6.0 minutes 5.7 minutes Expedited Delay per 100 TM 35.9 minutes 36.7 minutes 36.6 minutes 35.4 minutes Freight Delay per 100 TM 74.0 minutes 81.1 minutes 78.6 minutes 73.8 minutes Overall Delay per 100 TM 48.6 minutes 49.1 minutes 47.8 minutes 45.2 minutes

56 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations do not significantly influence passenger trains’ average speed, which stays consistent at about 36 miles per hour. The average speed of expedited freight trains (those with highest operating priority) slightly decreases with the added passenger trains operating over the network. However, when both sets of capacity improvements are implemented, average speed returns to the level seen in the base case. The same pattern and results apply to general merchandise freight trains. Overall, the average speed parameter increases throughout each simulation case. Delay Percent: This is delay time as a percent of pure run time. Compared to Simulation 1 val- ues, this parameter increases across all train types when new passenger services are introduced in Simulation 2. The highest increase is experienced by general merchandise freight. Track enhance- ments under Simulation 3 and 4 do appear to mitigate delay percent, but at different degrees: passenger trains’ delay decreases from 4.7% down to 4.2% which is still higher than in the base case. On the other hand both expedited and general merchandise freight ultimately experience better than base case performances; the same result is also reflected in the overall parameter. Delay per 100 Train Miles: This measure of system performance shows a trend similar to Delay Percent with respect to train types. Again, passenger trains experience higher minutes of delay once new services commence; and even if track improvements mitigate delays, performances do not improve or equate to original conditions. On the other hand, freight trains show better than base case performances under future build scenarios. In terms of overall delay per 100 train miles for freight and passenger trains combined, the results are better than the base case. 4.4.2.4 North Sound Operations Simulation Summary In line with previous simulation effort, this study shows that the addition of Bellingham- Everett regional rail service, plus the operation of one additional Cascade round trip Seattle- Vancouver, will not degrade current freight performance, but instead will improve it, assuming concurrent track capacity improvements. This round of simulations confirms the validity of the improvement package, whether or not the Bellingham-Everett regional rail service is established. The results do indicate minor increases in delay to passenger trains in Simulation 4 versus the base case Simulation 1. Mitigation of such delay may require operational changes or track capac- ity enhancements. As with the 2008 study, the 2010 round of simulations did not test any potential increased levels of freight service in combination with the added passenger trains. An increase in freight service levels could include coal and grain traffic for export from ports along the route. Importantly, BNSF advised that its review and comment on the RTC simulation did not con- stitute BNSF agreement to plans to implement a regional rail service between Bellingham and Everett. BNSF explained that its traffic patterns change over time, so baseline conditions will change. If the regional rail service were to materialize, BNSF said it will perform an independent operations simulation of the line to confirm system performance. 4.4.3 North Sound Grid Time Analysis 4.4.3.1 Introduction There are various levels of analyses that can be applied to determine whether the capacity of a line is adequate. A relatively simple grid time analysis can be used as a first cut to determine whether there may be a potential capacity problem in a corridor that warrants further investiga- tion. In this example, the BNSF rail corridor between Blaine and Everett was tested and evaluated as a conceptual level grid time analysis screening, albeit more detailed in terms of inputs than the LOSSAN grid time analysis as described in Section 4.2.

Best Practices 57 The reader should note two context elements for the grid analysis described below: • Grid time analysis would most naturally be the first and simplest scoping-level evaluation of service capacity for a rail corridor. For purposes of illustration the guidebook team chose an alignment that had earlier been modeled in RTC so that readers would be able to compare the products of each approach. • For purposes of simplicity the team is reporting the results of the grid time analysis for only a portion of the RTC-modeled service territory described above, namely the track segment between Everett and Blaine. In a simple grid analysis the capacity of a single track line to handle traffic is dictated by the time it takes a train to travel the distance between two adjacent passing tracks and clear the way for an opposing train (one-way grid time). The sum of the forward and reverse move grid times (turnaround grid time) for a balanced operation is the total time taken by a pair of opposing trains to cross a single track section. The single track section with the highest grid times is the most restrictive for the movement of the trains over the route between two major points such as between terminals and defines the maximum capacity for that segment. Ordinarily, such a “quick” analysis incorporates conservative assumptions and is easy to apply as a screening device. Normally if such application does not suggest any capacity problem on a line, then a railroad can be confident that capacity is sufficient and may not need to investigate further. However, if the grid time analysis shows that there is a potential capacity problem on a corridor, it is then recommended that a series of corridor simulations be conducted on a rigorous basis to test and evaluate the operations and capacity issues over the corridor. Based on the results of these simula- tions, adequacy of line capacity is then confirmed. In this grid time analysis, the theoretical capacity of portions of the route was calculated over each of the corridor single track segments (“grids”). The theoretical capacity calculation assumes the availability of an unlimited supply of trains at both ends of the line throughout the day. To correct for this oversimplification, a downward scaling factor is applied to derive the practical capacity of the line. The theoretical number is reduced to reflect conditions that all railroads encounter (i.e., track maintenance work, slow orders, unplanned disruptions, etc.) that reduce the effective capacity of the route. For this analysis, given the similar- ity of traffic on the line, the practical capacity of a line with one or more single track segments is assumed to be 75% of the theoretical capacity railroad design engineers seek to achieve. The grid time for a train is defined as the time taken from a stopped position at the switch at the start of the single track segment to train length distance (6,000-foot in this example) past the switch at the end of the single track segment (start of next siding or double/multi-track section). The train is also brought to a stop at the start of the next single track segment (end of siding or double/multi-track section). The train length distance allowance is applied to factor the train length, ensuring that the longest train has cleared the switch for the opposing train. By starting from stop, the additional time required for acceleration is accounted for. With the stop at the end of the passing track, the appropriate time loss for deceleration is applied as well. 4.4.3.2 Description of the Blaine–Everett Rail Corridor The Blaine–Everett rail corridor over the BNSF Railway is a low density freight operation that has in the order of 3 to 14 daily freight train movements per day plus 4 daily Amtrak passenger trains (two Cascades trains in each direction per day). In addition, daily light engine movements along with the occasional work train run across the corridor. The 88-mile line between Blaine and Everett is comprised of a single track with 8 sidings of over 6,000 feet for trains to meet one another. Current freight trains that operate over this segment include loaded coal trains in the northbound direction, empty coal trains in the southbound direction, and daily manifest trains in both direc- tions. Table 4-10 shows the siding names and lengths, locations, grid lengths, and average train velocity over each segment (from RTC simulations).

58 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations GRID BREAKDOWN Blaine - Evere Rail Corridor Siding length (feet) Siding Name Milepost (miles) Single Track Grid Lengths between Sidings* (miles) Average Train Velocity (mph) 4,602 Blaine 119.58 118.61 Between sidings: 1.81 40 8,588 Swi 116.8 115.1 Between sidings: 1.91 40 10,150 Intalco/Custer 113.19 110.94 Between sidings: 2.79 40 8,478 Ferndale 108.15 106.37 * Bellingham 98.07 97.11 * MP 96.73 96.73 96.36 Between sidings: 12.91 30 6,347 Bellingham Staon(South Bellingham) 93.46 92.20 * Samish 83.53 82.76 Between sidings: 11.30 35 8,884 Bow 80.90 79.06 4,635 Burlington Yard 71.91 70.36 Between sidings: 11.68 35 6,075 Mt Vernon 67.38 65.50 Between sidings: 8.97 50 6,381 Stanwood Staon 56.53 55.18 8.94 50 10,680 English Staon 46.24 43.9 * Marysville 39.19 38.69 Between sidings: 6.88 50 Delta Juncon 37.02 Evere Staon 32.00 * Note: Grid length (in miles) taken only between sidings capable of holding a train of 6,000  in length. The one excepon is Blaine, as all trains must stop there for customs inspecon. Table 4-10. Grid breakdown of the Blaine–Everett rail corridor.

Best Practices 59 The single track grid lengths range from approximately two miles in distance to as much as 13 miles. The longer the distance of each grid segment, the less capacity that the system is capable of handling. 4.4.3.3 First Pass—Grid Analysis for Freight Operations Only As a first pass, the grid analysis was performed over the corridor under the assumption of all trains having equal priority and all being freight trains. A standard freight train of 6,000-feet was used in this example as the basis of the conservative movement calculations. The train would start and stop at each siding, with the average speed as obtained from the aforementioned North Sound RTC simulation for the individual grids. An acceleration and deceleration time of 10 min- utes was added to each grid movement. At the northern three sidings, and due to their proximity with the border crossing, it was assumed that the effects of the border crossing delays would add 45 minutes of delay to each train in each segment, thus restricting the free flow of trains. As can be seen in Table 4-11, the grid analysis yields a relatively free flowing operation given a freight only environment, except for the problems related at the northern end due to border crossing delays. With current freight operations in the order of 14 daily movements and with a grid analysis that shows capacity between 26 to 50 trains per day capability for most of the route, there seems to be sufficient capacity. However, as this corridor is shared with passenger trains, the grid analysis is further refined to account for the prioritization in effect with such movements. As a general rule, the dispatcher must ensure that sufficient track capacity be available for high priority passenger train service. Accordingly, for each passenger train, the grid analysis must incorporate a total of three available grids to ensure that the train does not stop or get delayed due to freight interference. The individual grid is analyzed along with the grid ahead and grid behind, for a total of thee-grid calculation for each passenger train. 4.4.3.4 Addition of Existing Passenger Trains to the Mix As can be seen in Table 4-12, the practical capacity of each segment has been reduced to account for the prioritization of the existing Amtrak Cascade passenger trains. Other than the northern three sidings being affected by the border delay issues, the current 14 daily freight trains are within operational capacity limits of the existing infrastructure with a range of 21 to 46 trains per day for most of the route. 4.4.3.5 Addition of Future Passenger Trains to the Mix As Amtrak increases service by two daily Cascade movements per day (one additional daily train per direction) and the North Sound Regional Rail Service introduces four weekday move- ments (two weekday trains per direction), a further grid analysis was done to test the effects to capacity over the corridor, as seen in Table 4-13. With the introduction of a third Amtrak Cascade daily train and with the North Sound Regional Rail Service introducing two morning and two evening trains between Bellingham and Everett, the available capacity over the corridor becomes much more restricted. The grid analysis at this stage indicated that three segments south of Bellingham are at or close to their practical capacity limits, requiring further analysis (through simulations) to quantify the effects and develop appropriate remedies. The three questionable areas included the grids between: • Ferndale and Bellingham where the calculated practical freight train capacity was 18 trains per day but actual daily freight traffic was 14 trains; • Samish and Bow where the calculated practical freight train capacity was 16 trains per day but actual daily freight traffic was 14 trains; and, • Bow and Mt. Vernon where the calculated practical freight train capacity was 17 trains per day but actual daily freight traffic was 14 trains.

60 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations GRID BREAKDOWN Blaine - Evere Rail Corridor Siding Name Single Track Grid Lengths between Sidings (Miles) Total Time for Each Train to Cross Grid Including Border Delays (Minutes) Theore„cal Maximum Number of Freight Trains per 24 hour Period Prac„cal Maximum Number of Freight Trains per 24 hour Period (75% of Theore„cal) Actual Freight Trains per Day (Peak Days) Blaine 1.81 61.6 23.4 17.5 14 Swi 1.91 61.8 23.3 17.5 14 Intalco/Custer 2.79 63.1 22.8 17.1 14 Ferndale Bellingham MP 96.73 12.91 41.0 35.1 26.3 14 Bellingham Staon Samish 11.30 33.8 42.6 31.9 14 Bow Burlington Yard 11.68 34.5 41.8 31.3 14 Mt. Vernon 8.97 23.9 60.3 45.2 14 Stanwood Staon 8.94 23.8 60.4 45.3 14 English Staon Marysville 6.88 21.4 67.4 50.5 14 Delta Juncon Evere Staon Table 4-11. First pass grid analysis for freight operations only between Blaine and Everett.

Best Practices 61 Table 4-12. Impact of existing passenger trains. Siding Name Single Track Grid Lengths Between Sidings (Miles) Total Time Freight Slots Unavailable Due to Psgr. Train Priority (Minutes) Available Minutes in Day for Freight (Minutes) Theorecal Maximum Number of Freight Trains per 24-hour Period Praccal Maximum Number of Freight Trains per 24-hour Period (75% of Theorecal) Actual Freight Trains per Day (Peak Days) Blaine 1.81 83.7 1,356.3 21.5 16.1 14 Swi 1.91 100.5 1,339.5 21.7 16.3 14 Intalco/Custer 2.79 161.3 1,278.7 20.3 15.2 14 Ferndale Bellingham MP 96.73 12.91 276.8 1,163.2 28.4 21.3 14 Bellingham Staon Samish 11.3 303.9 1,136.1 33.6 25.2 14 Bow Burlington Yard 11.68 276.9 1,163.1 33.7 25.3 14 Mt. Vernon 8.97 194.5 1,245.5 52.1 39.1 14 Stanwood Staon 8.94 171.5 1,268.5 53.2 39.9 14 English Staon Marysville 6.88 128.4 1,311.6 61.4 46.0 14 Delta Juncon Evere Staon

62 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations In general, whenever the calculated capacity and the actual capacity numbers are relatively close to one another, a potential area of conflict is identified. In this case, it is simply a “flag” that is raised when there is an introduction of the new passenger services. A physical capacity constraint of some sort is likely to occur that will require some sort of physical capacity enhance- ment. By focusing the next level of analysis over the identified three segments, it is likely that the reduction of grid lengths may be the most appropriate recommendation (i.e., connecting sidings such as between Samish and Bow and between Bellingham with MP 97). 4.4.3.6 Impact of Additional Improvements As a final test of the impact of adding capacity improvements over the corridor, four sidings were connected making two longer sidings: Bellingham/MP 96.73 siding; and Samish/Bow sid- ing. These siding combinations have the effect of minimizing single track grid lengths between Siding Name Single Track Grid Lengths between Sidings (miles) Theorecal Maximum Number of Freight Trains per 24-hour Period Praccal Maximum Number of Freight Trains per 24-hour period (75% of theorecal) Actual Freight Trains per Day (Peak Days) Blaine 1.81 20.8 15.6 14 Swi 1.91 20.9 15.7 14 Intalco/Custer 2.79 19.0 14.2 14 Ferndale Bellingham MP 96.73 12.91 25.0 18.7 14 Bellingham Staon Samish 11.3 22.0 16.5 14 Bow Burlington Yard 11.68 22.8 17.1 14 Mt. Vernon 8.97 38.0 28.5 14 Stanwood Staon 8.94 39.5 29.7 14 English Staon Marysville 6.88 47.1 35.3 14 Delta Juncon Evere Staon Table 4-13. Impact of new passenger trains.

Best Practices 63 MP 106.37 (south switch of Ferndale siding) and MP 83.53 (north switch of new Samish/Bow siding). As can be seen in Table 4-14, potential freight train volumes increase dramatically on the remaining single track grid lengths in this mid-route area over previous volumes. It should be noted that this grid analysis assumed a lesser improvement than the RTC simu- lation regarding joining sidings at Bellingham. In RTC, the South Bellingham (or Bellingham Station) siding was joined with the Bellingham siding at MP 97.0. In the grid analysis, sufficient capacity could be provided by joining MP 96.73 siding with Bellingham siding; this is a sub-segment of the South Bellingham (Bellingham Station)/Bellingham siding improvement concept tested by RTC in Simulation 4. 4.4.3.7 North Sound Grid Time Analysis Summary The results of the grid analysis raised a cautionary flag following the potential introduction of a new passenger and commuter rail service over this corridor. Even though the existing opera- tions showed adequate capacity, this capacity was “lost” when a new passenger/commuter service Siding Name Revised Grid Lengths (miles) Theore cal Maximum Number of Freight Trains per 24-hour Period Prac cal Maximum Number of Freight Trains per 24-hour Period (75% of theore cal) Actual Freight Trains per Day (peak days) Blaine 1.81 20.8 15.6 14 Swi 1.91 20.9 15.7 14 Intalco/Custer 2.79 19.0 14.2 14 Ferndale 8.30 37.1 27.9 14 Bellingham / MP 96.73 2.90 52.9 39.7 14 Bellingham Staon 8.67 29.8 22.4 14 Samish / Bow Burlington Yard 11.68 23.5 17.1 14 Mt. Vernon 8.97 38.0 28.5 14 Stanwood Staon 8.94 39.5 29.7 14 English Staon Marysville 6.88 47.1 35.3 14 Delta Juncon Evere Staon Table 4-14. Impact of joining Samish and Bow sidings; extending Bellingham siding to MP 97.

64 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations was contemplated. However, the analysis did identify the area most in need of further study and set a path for subsequent evaluation with more sophisticated analytical tools. 4.4.4 North Sound Web-based Shared-use Tool Analysis 4.4.4.1 Introduction As a third test, the Web-based Freight-Passenger Rail Corridor Project Screening Tool, also known as the Shared-use (SU) Tool, developed for the National Cooperative Freight Research Program (NCFRP) Project 30, was used to determine if the capacity improvements planned for the Cascades would enable the implementation of two round trip, peak period North Sound regional rail trains on weekdays. The SU Tool was developed in response to a broad interest from public planning agencies in having access to a scoping-level instrument that would narrow down the range of potential corridor service opportunities and isolate locations that are worthy of further study. It includes a number of features, such as train performance logic, that are more sophisticated than the simple grid analysis described above. The SU Tool may thus be viewed as an intermediate level approach, falling between simple grid time analysis and the very precise (and data intensive) service capacity output features of RTC. The SU Tool requires MS Silverlight (a free web browser add-in). If this not already installed, the user will be prompted to download it. The SU Tool works on a browser that supports the Silverlight add-in. Per the SU Tool website, these browsers include Internet Explorer version 6 or later, Safari, Google Chrome, and Firefox web browsers. The SU Tool webpage is self-explanatory and a user manual is provided on the website to guide through the steps involved in creating a simulation model. The SU Tool organizes data pertaining to a simulation analysis (which may contain multiple scenarios) into a container system. The three principal data containers in SU are Folders, Rail Systems, and Operating Plans. • Folders contain all of the data required for an analysis. A Folder contains one or more Rail Sys- tems, which in turn contains one or more Operating Plans. Trains are contained in the Folder, and any train in the Folder may be referenced by any Operating Plan in any Rail System. • The Rail Systems container consists of track and related information (grades, curves and speed zones), all other infrastructure features, station lists, timetable routes, and traffic control blocks. Rail Systems include one or more Operating Plans. • An Operating Plan contains a list of selected trains from the Folder. Each train has a desig- nated timetable route and operating schedule. An Operating Plan also contains a set of central dispatcher parameters and other operating plan elements. The SU Tool provides two ways of coding the Rail System (network). There is a visual interface for beginners, which can be used to code railroad infrastructure (segments, switches, speed limits, curvature, elevation, and stations). For more experienced users, a table-based interface is avail- able to import data pertaining to infrastructure in a Comma Separated Values (CSV) file format. The table-based interface can be used to export the network files which are created using visual interface and recreate a new alternative with minimal effort, instead of building from scratch. For the North Sound simulation, an attempt was made to use the data interface to export the network from RTC simulation directly into the SU Tool using custom programming. Due the conversion issues, modelers were forced to resort to the visual interface. However, this effort provided a valuable lesson in understanding how the network is coded, and also significantly helped in coding the alternatives. The visual interface consists of click and drag methods to code the network; the details of cod- ing are explained in the SU Tool user manual. Once a segment is coded, the attributes (e.g., train

Best Practices 65 speed limit) of the segments can be modified by clicking the segment and choosing edit speed limit pop-up. The passenger and freight speed limits can be entered manually. The network was developed in incremental steps to check for any errors at each stage. Once the network is coded, a route file was created. Each route includes a “begin” and an “end” station, along with the inter- mediate stop locations. SU Tool provides visual tools to create trains. A train consist was created for each train using the number of locomotives and cars along with empties and loaded cars data provided by BNSF. Using the trains created, modelers developed a train timetable based on the route data from the network. The timetable includes arrival, departure, and dwell time at each location on the route. Likewise, an operating plan was developed using the train data. There are additional param- eters in the operating plan container, such as set central dispatcher parameters and other operat- ing plan elements. For simplicity, these parameters were left to default values. As with the grid time analysis, the basic train operating patterns for the SU Tool analysis were the same as defined for the foregoing RTC simulation description. Also, the goal of the analysis was the same. That is, do the improvements envisioned for more Amtrak Cascade trains provide sufficient capacity on the rail network between Blaine and Everett to host new North Sound regional rail trains and still maintain fluid, reliable freight and passenger operations? 4.4.4.2 Network Simulations As a part of SU Tool analysis three cases were simulated: • Simulation 1 (base case) represents existing track configuration between Everett passenger station to Bellingham and Blaine along with existing trains including two Cascade round trips between Seattle and Vancouver and BNSF trains. The analysis period was reduced to one day to reduce the simulation effort. The peak day was chosen such that it has maximum trains in a given week. • Simulation 2 has the same track configuration and freight trains as Simulation 1. However, with respect to passenger trains, it adds the two Bellingham-Everett weekday regional rail trains and a third Cascade Seattle-Vancouver round trip on a mid-day schedule. • Simulation 3 adds track improvements by extending the siding at South Bellingham north to Bellingham siding, joining Samish and Bow sidings, and implementing a universal crossover at MP 81.0. The trains were carried over from Simulation 2. 4.4.4.3 Simulation Results The SU Tool provides a summary of the train operational effects as changes in average speed and delay by individual train and by train type: freight and passenger. The SU Tool simulation results by train type are summarized in Table 4-15. A speed profile for the North Sound service appears as Figure 4-6. Case Average Speed (in mph) Average Delay (in minutes) Passenger Freight Passenger Freight Simulaon 1 (Base Case): No Build (Exisng Trains) 50.7 29.0 2.0 160.6 Simulaon 2: No Build (Future Trains) 48.3 28.5 15.0 145.9 Simulaon 3: Build (Future Trains) 47.2 30.5 6.8 52.4 Note: Simulaon 2 runs are unstable when simulated with different random numbers. Table 4-15. Simulation results from SU tool.

66 Capacity Modeling Guidebook for Shared-Use Passenger and Freight Rail Operations As shown in the table, Simulation 3 results (average speed and average delay) for freight trains were better than the Simulation 1 (base case) results. Improvement in the freight trains perfor- mance can be attributed to the infrastructure improvements in the Simulation 3 network. Also, the reduction in the average speeds and increase in average delays for passenger trains in Sim- ulation 3, when compared with the Simulation 1, can be attributed to the Bellingham-Everett weekday regional rail trains, which have more stops compared to the Cascade trains (Seattle to Vancouver passenger). The Simulations 1 and 3 runs behaved fairly consistent with different ran- dom numbers. The Simulation 2 results may not be reflected accurately as additional runs with different random numbers revealed instability, with significant delays and reduction in speeds for both passenger and freight trains. 4.4.4.4 North Sound SU Analysis Summary The results from the SU Tool analysis indicate that the addition of the two North Sound regional rail round trips and an additional Cascade round trip would have no negative impact on BNSF freight service, given the enhanced network assumed for Simulation 3. However, passenger train speeds and delays would worsen. 4.4.5 Comparison of RTC, Grid Analysis and SU Tool Results The North Sound RTC simulation, grid time analysis and SU Tool analysis provide results that are in some ways quite similar. The RTC simulation showed that, with the new passenger trains (Cascades and North Sound regional rail trains), passenger train performance would deteriorate somewhat, but freight train performance would be enhanced. This is the same finding generated by the SU Tool application. While the results like average speed per passenger and freight trains are different, the findings overall are consistent. The grid time analysis found that sufficient capacity exists today for the new passenger trains. But mid-route between Blaine and Everett, conditions get tight. Such a finding does indeed raise Figure 4-6. Train performance calculation generated by SU tool.

Best Practices 67 a flag over potential service issues, which both the RTC and the SU Tool confirmed: passenger train performance suffers. It is worth noting that grid time analysis can be used to quickly identify potential solutions for capacity constraints and test them—albeit not as definitively as either of the two simulation methodologies. Table 4-16 provides estimates of the time requirements for the three different analysis methodologies that were applied in the North Sound case study. Both the grid time and SU Tool analyses were for just the Everett-Blaine segment. The estimate for the RTC analysis reflects the simulation of the Scenic and Bellingham Subdivisions in addition to the Everett-Blaine segment, and thus is much higher. The estimates for labor hours are surrogates for estimated costs. The larger point of the comparison is that each analytical method has its place. Grid time analysis should be considered an up-front tool, used to understand whether or not a capacity constraint exists. Operations simulation, on the other hand, should be used to investigate the issue further, and to shape and define potential solutions. Approach Labor Hours Data Needs Comment Grid Time Analysis 40 Track charts, employee me table, and train counts for Evere to Blaine Quick for spong poten al trouble areas. Difficult to apply to complex networks. Web-based Shared-use Tool 120 Track charts, train opera ng paerns, and equipment detail for Bellingham Subdivision Good scoping-level tool. Lacks tools to visualize train simula on. Not user friendly to test network robustness, i.e., running mul ple scenarios. Model u lity in complex network situa ons not tested in this case. Data saved on web servers. Free tool. Rail Traffic Controller 400 Track charts, employee me table, train opera ng paerns, and equipment detail for Scenic and Bellingham Subdivisions Proven opera ons simula on methodology. Good tools for visualiza on of train simula on. Easily calibrated to reflect actual dispatcher prac ces. Runs on a local computer. RTC licenses must be purchased. Table 4-16. Summary of analysis methodologies for North Sound case study.