Integrating Aviation and Passenger Rail Planning (2015)

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.

98 Introduction and Structure Chapter 7 is presented in the following two parts, which follow after this introduction and review of major themes that are emerging from the research: • Part One examines the experience of the Regional Airport System Plan Analysis (RASPA) in dealing with issues of inte- gration with HSR planning in the region, and includes a discussion of the modeling tools employed (Figure 7-1). • Part Two deals with the physical site planning issues associ- ated with the most promising of the possible connections between the San Francisco International Airport (SFO) and the California High-Speed Rail alignment. The site planning case study provides a good example of how present design options may have been constrained (or even determined) by earlier physical infrastructure planning decisions. Major Themes Emerging from the Research • Some planners think they lack the tools to understand how passengers will divert from a specific airport. Managers at the Metropolitan Transportation Commission (MTC) believe people need to think bigger regarding modeling needs when considering diversion from air to HSR. • HSR was not explicitly modeled as a feeder mode for air because earlier studies concluded that the potential of HSR as a feeder was minimal and also because of challenges related to data collection and modeling. Several organiza- tions interviewed asserted a strong interest in evaluating air/rail complementarity. • Managers of the modeling process believe that the response of the airlines to HSR is unknown, challenging the resulting impact on the operations and the financial health of the airport. HSR will affect airports’ parking revenue, landing fees, etc., and it is unclear what this means for financial feasibility of airport bonds. • The models developed for statewide project development and environmental forecasting are not well equipped to support an in-depth exploration of policy analysis and scenario analysis. A scenario analysis of different fares is an example of the difficulties with the tools as initially devel- oped. Models which are slow to run pose an impediment to quick turn-around analyses appropriate for many policy questions. • In interviews with airport managers, a critical institutional issue was the set of possible actions available to an airport acting as an essential player in intermodal planning. The airport authorities were indirectly involved in RASPA, sitting on the oversight committee, but felt constrained in their possible actions. The airport noted its interest in having a broader institutional relationship, yet airports, and SFO specifically, are restricted by the FAA and local authorities. • In implementing a major new program like California High-Speed Rail, the site planning options are often fixed and difficult to change. Planners for the rail system at SFO are trying to figure out how to accomplish a one-transfer connection between the long-distance rail and the airport terminals, which may prove to be difficult because of existing infrastructure. Part One: The Regional Airport System Plan and HSR Overview of the RASPA The introduction of HSR service in California presents new inter- and intra-regional transportation opportunities. For interregional trips, it has the potential to reduce congestion in the aviation system and on the ground. For intra-regional trips, it presents a new alternative to driving and a new airport access mode. In an effort to define the role of rail service in the context of air service in the Bay Area, the 2011 Regional Airport System Planning Analysis evaluated the possibility C H A P T E R 7 The Role of Rail in Airport and System Planning in Northern California

99 of accommodating some air passengers on rail. The RASPA effort is an ambitious regional aviation system planning study for the San Francisco Bay Area, which is a joint effort between the Association of Bay Area Governments (ABAG), San Francisco Bay Conservation and Development Com- mission (BCDC), and MTC, under the direction of the RAPC (Metropolitan Transportation Commission 2011). The RASPA was managed by the Regional Airport Planning Committee (RAPC) made up of elected officials from the three regional agencies and staff from the region’s airports. The RASPA pro- vides both analysis and guidance on requirements for airports in the region; this is referred to by airports and the FAA when preparing airport master plans and environmental documents for proposed airport improvements. The results of the study are also used to plan surface transportation investments to provide access to airports by the MTC. In evaluating the potential of accommodating some air passengers on HSR in the Bay Area, the RASPA analysis uti- lizes existing modeling tools. The RASPA therefore represents a comprehensive study that showcases the ability of existing models to address air/rail integration. Evaluation of the RASPA provides the opportunity to examine existing tools that model how HSR and the aviation system will interact. High-quality intercity rail could be either an alternative mode for intercity transportation in a regional system, a feeder mode for the air network, or both. By discussing the process and outcome of the 2011 RASPA, one can observe how existing tools allow for analysis of air and rail integration. The focus of the RASPA case study is also institutionally unique, as the agency that is home to the RASPA also sponsored development of the main forecasting tool, the California High-Speed Rail Ridership Model (CHSRRM). This allows for the focus to be on the substantive content of the models, rather than lack of com- munication between separate agencies. Modeling Tools Executing the RASPA, or generally modeling the integration of air and rail, relies on the fidelity, scope, and power of model- ing tools. This section addresses existing tools—the current CHSRRM and a predecessor model—specific to modeling air and rail integration in the Bay Area. This section discusses their use, strengths and shortcomings, and additional tools that would greatly assist in modeling air and rail integration. Models of air/rail integration must address a broad set of questions. A key question is passenger mode choice and the resulting ridership—how the presence of rail will impact air ridership as both an alternative mode and a feeder mode. Along with ridership, though potentially not directly cor- related, comes the change in aviation operations due to the presence of rail. The changing landscape of transportation options will necessarily change the existing environmental footprint, necessitating environmental models. Stakeholders the Research Team interviewed in the RASPA process tend to agree that, as it pertains to the relationship between rail and SFO, rail could play a strong dual role as a: • Competitor to air, diverting passengers from interregional aviation travel; and • Collaborator with air, with rail acting as a feeder service to the aviation system. Modelers, regional planners, and SFO (as the main inter- national airport in the Bay Area) recognize the importance of this dual relationship. However, the potential of HSR to play this dual role is not consistently understood due the complexity of the subject, which poses challenges to the existing modeling tools, as discussed in the following subsections. Modeling HSR Ridership The CHSRRM was developed to meet the following three objectives (Cambridge Systematics 2007): • To evaluate HSR ridership and revenue on a statewide basis. • To evaluate potential alternative alignments for HSR. • To provide a foundation for statement planning purposes and for regional agencies to understand interregional travel. The CHSRRM was used by the RASPA to explore HSR as an alternative to air travel in order to manage existing aviation capacity. In order to appreciate the capabilities and limitations of the model for this purpose, it is important to overview its structure and calibration. Figure 7-1. Airports included in the northern California RASPA study. Source: RASPA.

100 San Jose and Los Angeles. This will affect main mode choice, drawing some passengers from the air mode. Other HSR traffic will be drawn from the highway mode, from changes (mediated through logsums) in destination choice because Los Angeles is a more attractive destination, and (again because of logsum increases) from induced interregional travel. If the HSR terminal in either San Jose or Los Angeles were made more accessible through either relocation or improved access/ egress linkages, the same effects would result, this time medi- ated through increases in the access/egress mode logsums for HSR. On the other hand, collocating a HSR terminal and San Jose airport would not have any clear impact, because HSR is not included as an access/egress mode for the airport. Critiques of the Model A number of potential issues with the CHSRRM were dis- cussed in a review of the model (Brownstone et al. 2010). In broad terms, the approach taken by the MTC/CHSR mod- elers includes a model development phase and a model vali- dation phase. In the model development phase, survey data were employed to estimate model parameters. The individuals surveyed were interregional trip makers, contacted through surveys; the mode choices of the individuals surveyed were not necessarily representative of California interregional travelers. For example, nearly 90% of long-distance (over 100-mile) business passenger trips are made by car, while 78% of the long-distance business travelers sampled for the study were traveling by air. It is important, in these circumstances, to adjust results obtained from the raw data to take into account differences between mode choices of the sample and the population. The methodology for doing this was recently shown to be theoreti- cally incorrect for the type of model employed by the modelers, though, at the time the study was undertaken the inaccuracy of this adjustment method was not known. The results, and the forecasts based on them (in particular of HSR mode shares), are therefore affected. There were other issues identified in the model development. Specifically the model structure does not allow for travelers to choose between HSR stations, thereby potentially exaggerating the importance of having frequent service at the single station that is judged to be “best” for a given trip. In the model validation phase, coefficients of the math- ematical model were adjusted so that it accurately replicated observed travel patterns in the year 2000. Model predictions of trips by mode, trip type, and passenger boardings on par- ticular air and rail routes were compared with observed values. Coefficients obtained in the model development phase were adjusted in order to obtain good agreement between predicted and observed values. As a result of this process, many of the model coefficients were assigned values that were considerably The CHSRRM consists of five components, which depict travel behavior as a sequence of choices: • Trip frequency (the most general). • Where to go. • What main (or line haul) mode to choose. • What access mode to use. • What route to take. Component models are developed for eight trip categories based on distance (over and under 100 miles) and purpose (business, commute, recreation, and other). The trip frequency model is multinomial logit, in which the household chooses to make 0, 1, or 2+ trips per day; utilities for the latter two options are constrained to be equal except for a constant. LOS variables, household characteristics, and location variables are included in the utility functions for 1 and 2+ trips. The service level variables include an interregional logsum that allows household responses to improved accessibility to interregional destinations by taking more trips. In essence, the logsum is a weighted average of the attractiveness of different destinations, based on the destination choice model. The destination choice is also multinomial logit. The utility of a destination depends on distance, area type, destination district, destination “size” in terms of employment, and number of households. Regional interaction terms, which change the utility of certain destination zones for households located in certain origin zones, are also included. Finally, the destina- tion choice model includes a mode choice logsum, a weighted average of the utility of different modes for traveling to the destination from the origin. Given that a household has chosen to make a trip to a certain destination, it then chooses the “main” mode for the trip. The modes for long-haul trips are car, air, conventional rail, and HSR. The model includes mode-specific constants, cost, in-vehicle time, service headway, and schedule reliability. Household characteristics, such as size and income, affect some of the modal utilities. The model also includes logsums for access and egress models. In this context, the logsum is a weighted average of the utility of different access/egress modes for reaching the main mode. The main mode choice model is nested, with non-car modes occupying the same nest. Two other nested models predict access and egress mode choices. Alternatives include drive/park, drop-off, rental car, and—occupying a single nest—taxi, transit, and walk/bike. Cost, in-vehicle time, and out-of-vehicle time capture the LOS for these alternatives. Distance, household characteristics, and airport dummy variables are included in the utility functions of selected alternative access and egress modes. It may be instructive to trace through how interactions between air and HSR are captured in this model framework. Suppose, for example, that a HSR link is added between

101 194,000 such trips revealed. In this analysis SFO has 2.4 mil- lion diversions from air to rail; ranking second in the total number of air passengers diverted was SJC with 1.9 million and OAK with 1.8 million Expressed as a percentage, SFO lost only 4% of its passengers to rail, with OAK losing 9%, and the smaller SJV losing 12% of its users to rail. It is worthy of note that this quantitative analysis was undertaken using a statewide rail ridership prediction model that did not include an airport choice submodel within it, and assumes that people are assigned to the airport closest to their trip ends. As described in some detail in the description of the analysis process, there are challenges associated with how the statewide model was applied in the regional RASPA planning process. Statewide diversion rates are being applied to the Bay Area; this contrasts to modeling diversion specific to the Bay Area. Additionally, from the calculation of the number of passengers who would divert to the HSR system, a calcula- tion of how many flights are eliminated is performed. Several stakeholders agreed that this is an indirect way of estimating the impact of HSR on the aviation system as it assumes the airline response is linearly related to ridership. An additional issue is that the model does not consider a system-wide over- all diversion, but rather adds diversion from certain markets. Thus, the analyst must select which markets to include. Diversion from Airports. In the vision held by some staff members the Research Team interviewed at MTC, HSR can be seen as functioning as a fourth regional airport. As the model applies the statewide diversion rate, the tools to under- stand how passengers will divert from a specific airport are not available. In this view, the planning process needs to think bigger regarding modeling needs when considering diver- sion from air to HSR. The current model involves a line haul choice and then an access mode choice; according to those the Research Team interviewed at MTC, these two could be more fully integrated, such that the modeling reflects that HSR is essentially a fourth airport. The lack of a station choice different than those obtained in the model development phase. In some instances, changes to the model coefficients were informed by professional judgments of the consulting team and the goal of replicating observed behavior. The resulting “validated” model, which is used to gener- ate subsequent HSR ridership forecasts, provides reasonably accurate “backcasts” for the year 2000, reflects certain patterns of behavior observed in the traveler surveys, and accords with professional judgments of the consultant. However, the combination of technical issues in the analysis methodology in the development step and subsequent changes made to model coefficients in the validation step implies that the forecasts could have large error bounds. Modeling Diversion from Air to HSR As stated in the objectives, the CHSRRM was developed to be a statewide model and generate statewide insights. The statewide ridership forecast includes estimates of the diversion of air passengers to rail. The number of air trips was forecasted in the No-Build case for HSR; this forecast was then compared with the forecast for a scenario that assumes a level of HSR service. The resulting change was the diversion of air travel to HSR in a given market (Gosling 2010). The calculated diversion rate was applied to the demand forecast for air travel for the RASPA study. As presented in Table 7-1, the RASPA analysis came up with some interesting interpretations of the model forecasts for the effect of HSR on air travel in the Bay Area. The strategic application of the CHSRRM to the Bay Area airports revealed diversion of 6.1 million air trips away from the three airports combined. As expected, the greatest amount of diversion occurs at SFO, the largest of the three airports, where about 2.4 million passengers were now assigned to rail. It is impor- tant to note that the RASPA study team was, in fact, able to make an estimate of how many of these trips assigned by the statewide model to rail were diverted from a connecting flight, rather than a trip with both origin and destination in California, with Source: RASPA Report, 2011 OAK SFO SJC Total Annual Passengers 20,655,000 64,356,000 16,305,000 O&D Passengers Diverted to HSR 1,776,000 2,218,000 1,935,000 Connecting Passengers Diverted to HSR 194,000 Undiverted Passengers 18,880,000 61,945,000 14,371,000 Percent Diversion 8.6% 3.7% 11.9% Table 7-1. 2035 diversion of Bay Area airport passengers to HSR in base case forecast–HSR initial phase, fares 83% of corresponding airfares.

102 for HSR to act as a feeder service is very low. The RASPA project dealt directly with the issue of diversion of trips from specific airports; however, according to those the Research Team talked to at MTC, there is little interest on the part of policymakers to explore in more detail the concept of rail as a feeder mode for air. This is because there is the belief that the potential is minimal. This tension is reported in several of the case studies in this report. In the Bay Area, it is believed by the stakeholders that SFO is the only airport where using HSR as an air feeder mode makes sense. Planners at MTC believe that use of HSR as a feeder mode at the smaller airports such as SJC or OAK will not prove to be significant, because of issues of location and scale. High- speed rail is not currently planned near OAK; other concerns at SJC include a weakness in long-distance flights. While the physical connection of HSR and air is easier at SFO with a short BART connection to the airport, it is still not seamless. The Millbrae intermodal station, which is nearby to SFO but not on the grounds, introduces this less-than-seamless connection (Figure 7-2). A traveler taking HSR to the airport will have to get to Millbrae and then connect to some kind of shuttle service to SFO; the design and site planning implications of this are explored in Part Two of this case study. The ridership impact of alignment changes related to a physical connection between rail and air in the Bay Area is difficult to address with existing planning tools. An SFO air- port task force was charged with looking at possible initiatives to improve the landside between air and landside modes. This task force considered encouraging the CHSRA to have the HSR alignment touching one of the terminals, with the assumption that the physical proximity would allow for a strong feeder service. However, it was not formally proposed because of the extreme expense; it was a “funding unconstrained” consider- ation. Some airport stakeholders argued that the costs and benefits of including a HSR station at SFO could have been more assertively explored. According to the Research Team’s interviews at both MTC and SFO, models capturing the ridership potential of rail as a feeder service for air could allow for more extensive scenario analysis. For example, such a model would optimally capture the potential for air/rail complementarity with the presence of joint fares—one ticket that includes the rail to air trip. The RASPA makes this recommendation and encourages CHSRA to pursue joint fares and joint baggage handling with air- lines. It would also be useful to have better tools to support the modeling of possible seamless connection between HSR and air, and evaluate the relative cost of options related to connections between HSR and the airport if they cannot be connected. The airport expressed an interest in a possible modeling tool to support scenario analysis about the physical transfer location and properties. component precludes this in the existing model, and further research into this aspect of future model refinement could be beneficial. For example, a model with airport/station choice integrated with airport ground access or station access mode choice might have allowed more scenario analyses (although that would have depended on the effort required to run the model for a given scenario) and could possibly have simplified the analysis process. The extent to which the results would have been more reliable is presently unknown. Modeling HSR as a Feeder Mode for Air In the ridership model, HSR was not explicitly modeled as a feeder mode for air. It was explained by the modelers that the possibility of rail as a feeder mode was excluded for two primary reasons. The first reason is grounded in the assertion by modelers that one tool cannot capture both competition and complementarity because of scope and underlying data. A different dataset would need to be collected to study com- plementarity compared with competition, including a survey of out-of-state travelers. It was deemed too expensive to do a separate survey of this separate market segment. The second reason was the earlier finding that air as a feeder mode has little potential, which observation in turn supports the first reason. This was discussed by the Cambridge System- atics modelers and reported by Brownstone et al. (2010). The diversion of connecting air travelers was included in prior modeling efforts, and it was found that this market segment is small; an earlier unpublished study concluded that this market segment accounts for less than one percent of HSR ridership and revenue potential. From a public policy point of view, it may be productive to undertake a systematic analysis of the possible roles of rail to serve in the feeder function for longer distance flights, as was advocated in the Research Team’s meetings with key managers of San Francisco Inter- national Airport—the persons who commissioned the earlier unpublished study. Given that the stated preference process reflects commonly held beliefs, the recent history of success- ful rail services in other parts of the world might result in a somewhat different set of perceptions. In wanting to better understand the potential for longer distance rail to widen their effective catchment area, SFO officials were acting in a manner consistent with the managers at the Copenhagen Airport, who accomplished exactly that in a short implemen- tation period. While many analysts remain skeptical about this submarket for rail, others strongly argue for it, as reflected in the estimate by SNCF that their investment in the Midwest could reduce connecting air passenger miles by 3.8 million, as reported in Chapter 6. Dealing with the Issue of Rail as a Feeder Mode. There are policy implications of the early finding that the potential

103 related to how air and HSR would exist in a market together. Additionally, there is much interest related to how the aviation modes will compete in a market with HSR. If air is compet- ing with HSR, how will airlines compete with frequencies and fares? From the rail standpoint, there is acknowledgment and agreement that the airlines will respond; however, there is dis- agreement as to how they will respond. Airlines are private companies and it is not in their business interest to be forth- coming. The airport operational profile that results from the introduction of HSR is unknown and could have positive and negative impacts on the airport. Capacity at SFO may open up if airlines reduce their intra- California service due to competition, or if airlines reduce their feeder aircraft service because of complementarity pos- sibilities. SFO’s long-term plan is in accord with this possibility. The airport is shifting focus to international traffic and long- haul domestic traffic away from short-haul domestic traffic. The urgency in dealing with the efficiency of operations is further underscored by SFO’s standing as the West Coast’s worst airport in on-time performance in 2010. Interestingly, these delays are in part due to efforts to increase short-haul domestic traffic in the earlier part of the first decade of the 2000s. With this increase in traffic, the airport is seeking all possible alternatives for preserving regional mobility while focusing more on international and long-haul domestic traffic. As part of their fiduciary responsibilities, airport managers remain concerned about any possible major reductions in air services, (from whatever motivation) which might impact basic sources Policy Scenario Analysis According to the managers at MTC, existing modeling tools were not designed to support wide exploration of issues and scenarios. One of the main issues, as discussed by a member of the RASPA analysis team, is that the CHSRRM demand forecast uses zone-by-zone estimates that underpin the model. However, some early applications of the model for policy purposes use a fixed matrix, which presents strong challenges to scenario analysis because—in practice—this matrix was not updated when a demand forecast with different parameters was estimated. An example of how this static matrix presents an issue is related to a scenario analysis of different fares. For example, in December 2009, new ridership estimates came out in the California High Speed Rail Business Plan (California High- Speed Rail Authority 2009), where the fare was assumed to be 83% of air. The ridership estimates were an update to the previous estimates that assumed HSR fare to be 50% of air. The zone-by-zone estimates were not re-estimated with the fare percentage change; rather, an updated zone-by-zone matrix was estimated offline using scale factors. While this represented a cost effective approach to model utilization, it is an example of a methodological decision having been made to deal with the realities of the complexity (and time consumption) of the main model. All stakeholders were interested in the lack of understand- ing about the response of the aviation sector in terms of an area of policy scenario analysis. There was extensive discussion Location of proposed long- distance rail station at Millbrae Distance to major domestic check-in terminals at SFO = 7,900 feet Figure 7-2. The proposed rail station is 1½ miles from the domestic terminal area at SFO.

104 related to model transparency and ease of use, performing such scenario analysis would be difficult. There are also issues with policy analysis due to the complexity of the demand model. The existing model is very complex and takes a good deal of computing power. This model is difficult to run, takes a very long time to process results, and also involves a high level of specificity. According to the stakeholders consulted, there is a need for a model that is more transparent and less computationally complex, and more flexible. Integrated Environmental Model An integrated ridership-diversion-environmental model would be extremely helpful, according to those the Research Team interviewed at the MTC. If HSR diverts a certain number of passengers and reduces flights by a certain amount, there could be a measurable change in environmental impact. This is particularly the case with noise and the change in the noise profile. HSR may have a noise benefit because flight operations are reduced, but others argue that HSR really just moves the noise because other people can then hear the new noise. In the Bay Area, there are residents on the peninsula in a new noise region, but it is unclear what the new HSR noise con- tour would look like. Tools are needed to study diversion and the implication of diversions, including noise, greenhouse gas emissions, and other pollutants. Institutional Issues A critical institutional issue is the set of possible actions available to an airport to be a player in intermodal planning. A key focus area of the RASPA is on the reduction of demand at SFO through airport diversion, intermodal diversion, or demand management. The airport remains strongly in support of HSR because diversion to HSR directly supports their plans to focus on long-haul domestic and international transportation. How- ever, the airport states that the potential of HSR to support their plans is unclear because of the lack of clarity of results from the analysis process. Given that lack of clarity, the airport also maintains some skepticism about the real potential of diversion to HSR. The airport and other stakeholders expressed concern that there will be some passenger diversion to HSR but that airlines will not shed their operations, simply operate smaller planes keeping frequency constant. They noted that all of this must be seen in the context of the question of political uncertainty that has surrounded the HSR program. Institutionally, the airport stated that the official dealings of HSR fall under the auspices of the CHSRA. Therefore, they leave the parameters of HSR up to CHSRA and do what they can to facilitate air/HSR interaction. of income, which range from landing fees to parking lot rev- enues. Airport managers feel they must be ready to examine all impacts of changes in the operational status quo. An additional issue related to ridership diversion from HSR to air is that airport planning is not driven by rail planning. The airport asserts that its plans cannot be dictated by activity in another mode; therefore, the San Francisco airport authority is not directly accounting for HSR in their master planning process. Their most important aspect of master planning is aligning their terminal capacity with their airfield capacity. Another challenge is that the airlines have a very short planning horizon, on the order of 2 years. This does not align well with the long-term planning of HSR. Policy scenario analysis related to changes in the air trans- portation system is something that the MTC analysts desire. There is currently no feedback related to airport operational performance or delays. The current aviation systems model cannot incorporate the idea that as SFO delays increase, pas- sengers divert to the other airports. The same issue is raised for HSR. If airport delays increase, passengers may divert to HSR, which needs to be incorporated into a future modeling process. Additionally, one of the underlying assumptions in the CHSRRM demand model is that HSR is a more reliable mode of transportation, because it is not affected by fog or weather, and passengers have certainty that the train is going to make it on-time as opposed to air travel. Alternatively, it would be important to include feedback due to higher air transportation reliability—possibly due to the implementation of advanced air traffic management technologies through NextGen—in the models of diversion to HSR from air. Furthermore, airports’ capacity benefits could be included into a future integrated model. A critical scenario that needs to be investigated from the airport capacity side is that diver- sions to HSR do not cause airlines to reduce operations, but simply reduce the gauge of their aircraft; in such a scenario, airport revenues go down, and capacity does not go up. The CHSRA ridership study assumes “the hassle and time variance of getting a boarding pass, checking luggage, and getting through security requires arriving at the airport ear- lier than at the train station without security checkpoints” (Cambridge Systematics 2007). HSR, on the other hand, will not have “elaborate security check-in procedures”; boarding passes will not be required to wait for a train; and seats are unassigned (Cambridge Systematics 2007). Therefore, under this assumption that the HSR will never have the same TSA security requirements as air transportation does, passengers do not have the stochastic security processing issue and the wait time is assumed to be 15 minutes. According to planners at the MTC, it was asserted that this would be a tremendous ridership advantage. However, it is uncertain whether security will be a small or a big issue; and, if this erodes, then some of the advantage of HSR disappears. However, due to challenges

105 imize the complementary role of rail with air. When viewed in terms of the competitive role for rail, it is often in the best interest of the rail operator seeking to minimize travel times to have no airport station stop whatsoever, as discussed in Chapter 3. By way of extreme example, the highly competitive travel times between Boston and Washington, DC, proposed by Amtrak in its “Next Generation Rail” proposal for the Northeast Corridor are based on a “Super Express” concept. This concept would stop only in New York City and Philadel- phia, bypassing all other stations; such a concept is designed to create a product that would be highly competitive with the major air markets in the area. More realistically, the rail planner seeks a single station with significant intermodal connections, including parking priced for the intercity rail market, and local feeder services in place to serve local demand. By contrast, the optimal rail station location for the com- plementary role would be within the air terminal area itself. Absolute minimization of the connecting times between rail vehicle and aircraft is desirable for the extreme case in which short distance feeder air service is eliminated and replaced by rail access; it is also desirable for the more common situation in which the traveler is offered both feeder air and feeder rail services to a major airport. The case study of the relation- ship between SFO and the California High-Speed Rail project reveals just how difficult it is to retrofit existing infrastructure to accommodate newly planned and highly desired services. Application of the Site Planning Concepts Consistent with the typology developed earlier in this report, the site planning options at the SFO can be viewed from an international perspective. The review of airport sta- tion locations in Europe and the United States suggests that the three major categories for airport rail station are relevant in the review of air/rail connections at SFO. Major infrastructure decisions were made at the time of the construction of the present BART extension to a point adjacent to the new Inter- national Terminal; during that process, many combinations of rail and people-mover options were examined. Thus, at present, the BART extension to a point near the International Terminal is in place as an elaborate “Y” connector in and out of the terminal area, as shown in Figure 7-3. Present/planned Services The site planning process must now take into consideration that, while the BART connection is in place, there is still a need to improve the connections between the airport’s four passen- ger terminals and both long-distance high-speed rail services, and regional rail services over the present Caltrain system. That system currently operates as a diesel commuter rail system to San Jose and beyond, with 77 mile corridor and 32 stations. Conclusion: Regional Airport Systems Planning Issues The Bay Area case study reveals that, despite the propitious circumstance of having the same agency responsible for devel- oping a detailed HSR demand model and the Regional Airport System Plan, the former was of limited value to the latter. In part, this reflects limitations in the demand model, particu- larly, its inability to address the role of HSR as a complement to the air transportation system and to explicitly model airport/ station choice. In other cases, capabilities of the model were not used to their fullest extent, such as in the assessment of the sensitivity of demand forecasts to different HSR fare scenarios. There is also a lack of tools to analyze the complementary mode. For example, it would be desirable—if very difficult— to predict how diversion of traffic to HSR, and the presence of HSR as a competing mode, will affect airline schedules. Lack- ing such a tool, planners often assume that flight frequency would change proportionally to passenger traffic. Institutional limitations also played a role. While planners recognized that HSR could create new opportunities for air- port demand management, there remain institutional and policy barriers to realizing these opportunities. This limits the application of existing tools to develop demand management strategies suitable for a multimodal environment and reduces the impetus to develop new ones. Overshadowing the RASPA analysis related to HSR was the great uncertainty about whether and when the system would be built. Given this uncertainty, there was little incentive to perform anything beyond the application of a very detailed model for analyses it was not designed to perform. While the tools could have been improved, it was perceived, with con- siderable justification, that investing in their development could not be justified at the time of the study. Part Two: Physical and Site Planning Issues at SFO Part One of this Chapter has established that the man- agers of the SFO airport see the potential for HSR services to complement their overall strategy to emphasize long- distance (and especially international long-distance) flights in the allocation of the finite capacity of the airport infra- structure (Reisman 2012). Several years of studying physical expansion options, including filling in portions of the Bay, have resulted in adopted policies to make better use of the existing airport plant and infrastructure. The question now turns to the ability of the managers of both HSR and the air- port to maximize the quality of the interconnection between ground services and aviation services. To review a basic theme of this ACRP report, the provision of high-quality transfer between rail and air can be used to max-

106 It currently operates 92 trains per weekday, and carries about 50,000 per weekday (Caltrain 2014). In development of its updated Strategic Plan during 2014, the railroad is committed to developing what it calls a “HSR/Caltrain Blended System,” to deal with near-term electrification of the line. This will have major implications for the quality of the connections between SFO and the counties immediately to the south. Travel times between Millbrae and San Jose will improve to 21 minutes for the non-stop operations, and 25 minutes for a one stop service. This “blended system” concept will, first of all, improve the ability for SFO to increase its rail catchment area to the south. The alignment is shown in Figure 7.4 In addition, the electrification will improve service to the North, to the Market Street/Embarcadero area. At present, the Figure 7.3. Location of BART tracks in and out of SFO. Imagery © Landsat, Google Earth. Figure 7.4. Proposed alignment for “blended service” between Transbay Transit Center and San Jose. Source: CHSR Authority, 2013.

107 east of the International Terminal, with walking connections to it and to the three terminals on the loop. However, no such an alignment for the HSR right-of-way is currently under consid- eration by the High-Speed Rail Authority, which is examining staged construction options that would improve the existing two-track Caltrain alignment in the area, with early electri- fication. Airport planners told us that the presently planned investment in Caltrain could change the planning context, with a new downtown terminal complex, and with “faster speeds and higher frequencies which alone would justify an on-Airport stop even if California HSR is never built” (Panel communication 2013). Thus, it is not clear how many tracks would optimally be diverted into such a tunnel, or which rail markets would be served. The emergence of the blended con- cept of electrified services from Caltrain and CHSR makes the combination of potential markets even more challenging. Site Planning Concept #2: People-Mover Shuttle to Rail Alignment: Build a Separate Airport Rail Station on the Rail Alignment at a Point As Close As Possible to the Airport Passenger Terminal Complex This kind of connection, as exemplified by the existing air- port rail station connection at Newark Liberty International Airport, would result if planners could build a new station parallel to the present rail alignment, getting the longer dis- tance rail passenger as close to the airport as possible, with as short a people-mover connection as possible. The fact that these users could get short distance people-mover access to all four air terminals with only one transfer is a positive char- acteristic of this site planning concept. Negative aspects of such a solution for the HSR station location include the fact that the existing Caltrain alignment is directly adjacent to several residential neighborhoods, as is evident in Figure 7-3. The aerial photograph in Figure 7-3 also shows how the BART rail tracks between the airport inter- change complex and Millbrae station have been placed in a cut and cover tunnel to minimize impacts on these residential areas. Fitting any major new transportation facility around the highway and the residential neighborhoods would be challeng- ing from an environmental perspective. Indeed, assuming that the new station were to require direct highway ramps, fitting all the needed infrastructure into the existing site would be nearly impossible. Site Planning Concept #3: Connect to Network at a Central Place: Connect with the Best Consolidated Rail Transfer Point Possible At the moment, the air/rail station planning by CHSRA is most similar to Site Planning Concept #3, in which the fastest Caltrain travel time between Millbrae and the down- town rail terminal is 18 minutes with current diesel propulsion equipment. The investment in electrification currently under environmental analysis will increase acceleration and decelera- tion speeds, with travel time improvements to 13 minutes, and the ability to run 6 trains per peak hour over existing track (Caltrain 2013). By way of comparison, these planned travel times from Millbrae could be similar to that currently offered on the Heathrow Express (15 minutes) between Heathrow and Paddington Station in London, which is provided four times per hour. In downtown San Francisco, the Caltrain line is planned to be extended by 1.3 miles to a new Transbay Tran- sit Center, located immediately south of Mission Street with connections into the Montgomery BART station on Market Street. Present in-vehicle rail travel time between BART’s Air- port station and its Montgomery station is 30 minutes, with 11 stops. Caltrain is now planned to operate 13 minute ser- vice into the lowest level of the Transbay complex, as part of a 6 track joint station with HSR. Funding is currently being sought for the new Caltrain terminal facility (Transbay Joint Powers Authority 2014). Site Planning Concept #1: Full Integration at Airport: Reroute the Long-distance Rail Line so That the Air Traveler Can Walk from the Train to the Check-in Terminal In this option, the SFO would operate like airports in Zurich or Amsterdam where the trunk rail line is placed in the “basement” of the airport terminal. From a historical perspective, it can be noted that the original designers of the present domestic terminal area of SFO did, in fact, make pro- visions for some kind of rail to go under the area generally defined as Terminals 1 and 3. While an easement was origi- nally left under the present central domestic garage, it was not anticipated in the 1970s that the airport would be served by both the regional BART trains and statewide rail trains. Con- struction of such a rail station potentially accommodating four (or more) tracks underneath a working airport would have posed a major challenge. While finding an optimal “Zurich style” alignment would be difficult (and expensive) airport planners at SFO note the desirability of a close physical connection from rail to air, preferably by walking which they described as the “inter- national standard” (Panel communication 2013). They note that enabling a walking connection would require a diversion of the rail alignment directly into the airport terminal area, preferably operated as a through service rather than as a stub- end terminal. Just where such a rail tunnel would be built is beyond the scope of this research. In previous planning efforts such a direct alignment was explored generally to the

108 mover (which loops around the original terminal complex) to Millbrae. Official statements from the Authority refer to some form of shuttle service on the existing BART track between Millbrae and SFO. The Research Team has not found any offi- cial agency reference to the option of extending the existing airport people-mover to Millbrae. Previous studies of rail stations in the area explored build- ing a 1.2 mile spur off the existing airport people-mover from the present BART station to Millbrae station. Alignments between the two points might either parallel South McDonnell Road (generally north of Route 101), or parallel the existing BART alignment (generally west of Route 101). The distance to Millbrae station is similar, but slightly longer than the distance of the existing people-mover to the consolidated car rental facility. If a major investment were to be made in a people-mover between SFO and Millbrae station, another design option which illustrates the logic of Site Planning Concept #3 could include using Millbrae station as the only rail transfer point for the airport, with present BART service to the airport routed only to Millbrae. If all six presently planned Caltrain peak hour services were to stop at such a newly expanded station, for example, combined northbound Millbrae service to Down- town San Francisco (i.e., Montgomery/Transbay stations) could offer 14 peak hour trains, with eight on the BART track and six on the Caltrain track, with a composite average headway of shorter than 4.5 minutes; this would be one of the most frequent airport-to-downtown rail services offered in the world. Whatever additional service frequencies might result from later long-distance investment on the Caltrain/ HSR alignment would further increase the diversity of services and destinations offered to the transferring air traveler. airport rail stop is located to access the maximum amount of connecting services and other intermodal facilities. As shown in Figure 7-5, the present Millbrae station location provides existing infrastructure (including a major parking garage) to support interconnection between HSR, Caltrain regional rail, BART and an extensive local bus feeder sys- tem. The station is located in a community center consis- tent with local public policy to encourage “transit oriented development.” BART Connections Between SFO and Millbrae The challenge for local designers to make the Millbrae transfer site work for air/rail complementarity is to provide high-quality connections between the rail station and the four air passenger terminal buildings at SFO. BART at some times in the past has offered direct service between SFO and Millbrae, and at other times has not offered it for many hours of the day–reflecting both operating and budgetary constraints. Even assuming 15/20 minute service between the two points for most of the day, which is consistent with good inter- national ground access practice, they are not consistent with people-mover schedule patterns within an airport complex. Even if additional rail shuttle services were added between the regional BART trains, most travelers making the long- distance rail-to-airplane transfer would need to proceed from the International Terminal to terminals 1, 2 and 3, whether by existing people-mover or by foot. In one interview conducted for this case study, a former CHSRA official commented that the best solution for the area would be the extension of the present airport people- BART Millbrae Station Figure 7.5. The present BART/Caltrain station at Millbrae is about 2,000 feet from airport property. Image © 2014 Landsat, Google Earth.

109 Summary: Site Planning Challenges The interface between the California High-Speed Rail alignment and the passenger terminal area of San Francisco International Airport represents one of the most important American case studies in the examination of a complemen- tary role for air and rail. Site planning challenges include the potential to improve airport access utilizing new combina- tions of both the multi-stop BART trains, and limited stop express Caltrain in the metropolitan market, in addition to consideration of several possible roles for the “HSR/Blended Caltrain System” for longer distance rail markets. In terms of air markets served, SFO represents a highly successful hub operation in which the airport must assemble passengers for important longer term flights. Indeed, SFO is the primary location for Northern California to remain competitive in the long-distance air international market. Logically, rail could play some role in providing access to those long-distance flights. Given that CHSR will not serve LAX directly, and given that runway conditions at San Diego constrain the development of international flights, SFO would be the strongest candidate in California for a symbiotic relationship between air and longer distance rail. Given the decision to focus major investments in the central portions of the HSR system first, California policy makers have the opportunity now to review their options concern- ing the connections to SFO. Methods to bring the future rail tracks into a tunnel under the SFO passenger terminal area could be reviewed, consistent with the logic of Site Planning Concept #1. Alternatively, Site Planning Concept #3 could emphasize methods to maximize the quality of the transfer between the four air passenger terminals and the Millbrae intermodal center. The planning process needs to continue, given that California will invest immediately in the Caltrain right-of-way. This will allow near-term design efforts to be informed by a better understanding of longer term needs for air/rail complementarity, involving metropolitan, regional and longer distance roles for rail at one airport. This planning process could acknowledge concepts devel- oped at the conclusion of Chapter 3, which emphasize the importance of understanding the needs of all stakeholders. Specifically, Chapter 3 concludes that often it is not in the interest of the long-distance rail operator to have additional stops for its fastest trains. If an airport stop is to be imposed on the long-distance system, it is logical to design that stop to serve as many markets as possible. Planning now underway for a major Millbrae station is consistent with the logic of Site Planning Concept #3. Optimally, all those involved in these decisions could be supported by demand modeling tools and data-sources that can help determine the importance—or lack thereof—of good connections and services for those using longer distance rail to access long-distance flights at SFO. Bibliography Bay Area Rapid Transit (BART). 2008. “BART to SFO Ridership Jumps 65%.” Available at http://www.bart.gov/news/articles/2008/news 20080626.aspx. (As of March 16, 2011). Brownstone, D., M. Hansen, and S. Madanat. 2010. Review of “Bay Area/ California High-Speed Rail Ridership and Revenue Forecasting Study.” Research Report UCB-ITS-RR-2010-1. California High Speed Rail Authority. 2008 (November 7). “High Speed Train Systems.” Retrieved from http://www.cahighspeedrail.ca.gov/ Business_Plan_reports.aspx. (As of Spring 2011). California High Speed Rail Authority. 2009. “Business Plan: Decem- ber 2009 Business Plan Report to the Legislature.” Retrieved from http://www.cahighspeedrail.ca.gov/Business_Plan_reports.aspx. (As of Spring 2012). Caltrain. 2013. “Peninsula Corridor Electrification Fact Sheet.” Available at http://www.caltrain.com/Assets/Caltrain+Modernization+ Program/Documents/Peninsula+Corridor+Electrificaton+Fact+ Sheet-2013.pdf. (Accessed March 2014). Caltrain. 2014. “Caltrain Strategic Plan Update.” Available at http:// www.caltrain.com/Assets/_Planning/Strategic+Plan/LPMG+ SP+jan+2014.pdf. (Accessed March 2014). Cambridge Systematics, Inc. 2007. “Bay Area/California High-Speed Rail Ridership and Revenue Forecasting Study.” Prepared for the Metropolitan Transportation Commission and the California High- Speed Rail Authority. Gosling, G. 2010. “High-Speed Rail Scenario Passenger Diversion.” Available at: http://www.regionalairportstudy.com/library/RAPC- High-Speed-Rail-Diversion-Memo-Final.pdf. (As of Spring 2012). Metropolitan Transportation Commission. 2000 (September). “Region- al Airport System Plan, Update 2000, Final Report.” Available at: http://www.mtc.ca.gov/planning/air_plan/RASP_FinalReport.pdf. (As of Spring 2012). Metropolitan Transportation Commission. 2011. “Planning.” Available at http://www.mtc.ca.gov/planning/air_plan/index.htm. (As of Spring 2012). Metropolitan Transportation Commission. 2011. “Regional Airport System Planning Analysis.” Available at http://www.regionalairport study.com/library/RASPA-2011_update/Volume_1-Sept_2011_ RASPA_Final_Report.pdf. (Accessed March 2014). Reisman, W. 2012 (April). “High-speed Rail Could Free up Valuable Space at SFO.” San Francisco Examiner. Transbay Joint Powers Authority. 2014. “Transbay Transit Center.” Available at http://transbaycenter.org/project/program-overview. (Accessed March 2014).