Electrical mobility : electric vehicles and electric bikes

Abstract

Introduction

The increasing complexity of social, political and economic factors associated with metropolitan planning have driven researchers to develop a new regime of modeling tools called integrated transportation and land use models (ITLUM). These tools are being developed in part based upon the recognition that transportation and land use planning is a complex process; the need to analyze various inter-related policy initiatives to support decision-making; and the requirement to satisfyingly involve stakeholder groups in the metropolitan planning process. The ITLUM tools being created are often complex in nature requiring large amounts of detailed data, resources (time and money), functionality, and expertise. Unfortunately, many of the more complex tools are not easily accessible by many planning agencies in the U.S. And, researchers often criticize the use of simpler modeling tools given the complex nature of urban systems as being too simplistic [1].

However, there is some evidence to suggest that a complex modeling approach is not necessarily appropriate for all aspects of the decision-making process and that a simpler tool may be sufficient in order to capture the complexities associated with understanding the transportation and land use system dynamics [2]. In fact, if a goal of the planning process is to better engage stakeholders, then creating ITLUM tools that are better accessible to stakeholders and decision makers alike may be a useful exercise [2]. Thus, the motivation for this research is to further explore how a simplified modeling approach (one that requires less data, resources, functionality, and expertise to operate[footnoteRef:1]) can support decision making within the context of the metropolitan planning process in the U.S. The question of concern for this research is focused on what are the opportunities and limitations of using readily-available data sources to populate, calibrate, and validate a simplified ITLUM tool and what are the implications? [1: The distinction between simple and complex modeling approaches are based upon four requirement categories: data, resources (time and money), functionality, and expertise. Simple modeling approaches have lower requirements and complex modeling approaches have higher requirements.]

Literature Review

Computer modeling tools have a rich evolutionary history and researchers have described the evolution in different ways. Miller, et.al., and Wegener describe the evolution from simple to complex with the goal of developing tools that are, generally speaking, more complex in nature [3], [4]. Mile, et. al., however, describes it slightly differently where tool selection to support transportation planning is a trade-off between system complexity and spatial complexity [5]. Most recently, Hardy, et. al., in discussing the application and use of transportation models for work zone design and evacuation modeling, describe the evolution in terms of a spectrum of modeling tools where selection is based upon five aspects: functionality, results, time, training, and cost [6], [7]. What is absent from these frameworks of model evolution and selection is an inherent decision-making functionality that a computer modeling tool supports. Today, decision makers rely on a spectrum of modeling tools to support the necessary decision-making categories associated with the metropolitan planning process.

The use of computer modeling tools has been an integral part of metropolitan planning and supporting the decision-making categories, though very few tools exist to look at the transportation and land use system in an integrated manner [8]. Within the transportation planning process an entire set of transportation forecasting modeling tools has been developed that date back to the 1950s and are primarily based upon the four-step planning process [9][footnoteRef:2]. Today, every federally designated metropolitan planning organization employs some type of transportation forecasting model. The land use planning process, however, is not as well-established in terms of tool development or application. Of the 35 largest metropolitan areas, only 12 were using commercially available land-use models [10]. [2: The four-step planning process consists of trip generation, trip distribution, mode split, and route assignment.]

Iacono, et. al, refer to the early development of computer modeling tools as being focused on the available expertise associated with developing a modeling approach and the availability of data to test the approach [9]. Over time, expertise began to grow and the ability to reasonably collect data increased such that the focus of model development shifted to incorporating more functionality into the model and better representation of real-world systems. For example, the development of microsimulation modeling approaches occurred in parallel with increasing computing power and data storage.

While many resources have been devoted to the development of more complex computer modeling tools (e.g., UrbanSim, TRANSIMS, SHRP II) less attention has focused on simpler modeling approaches. Just as one cannot use only a hammer to construct a house, one modeling tool cannot support the entire metropolitan planning decision-making process. Recently, the use of system dynamics[footnoteRef:3] as a simpler approach has been discussed as a means to support the metropolitan planning decision-making process. Abbas and Bell articulate twelve reasons why system dynamics might positively contribute to the metropolitan planning process [1]. Many of these reasons have been supported by later research most notably that of Sussman, et. al, who articulate the need for system dynamics as part of a regional strategic planning process to better engage stakeholders [2], [11]. [3: System dynamics is a modeling approach designed to incorporating complex feedback relationships to better assess system analysis. A detailed discussion is presented in Section Error! Reference source not found..]

To this day, researchers and policy makers continue to debate the role and purpose of computer modeling tools to support metropolitan planning. On one hand, researchers cite the need for complex, large-scale modeling tools that are able to include more functionality with a higher degree of accuracy in the results [12], [13]. These researchers suggest the need to further develop the complex and comprehensive tools that were created as a result of ISTEA and the 1990 Clean Air Act Amendments (CAAA) such as UrbanSim or TRANSIMS [13]. On the other hand, some researchers and decision makers indicate the high cost, difficulty of use, and large data requirements required to run large-scale models and identify the need for simpler models [14]. This group suggests there is a need to incorporate the complexities of the dynamic urban process by way of simpler computer modeling tools that are more accessible to decision makers [11]. The research presented here builds upon this debate and provides evidence to suggest that the use of a simpler modeling approach can be a cost-effective approach to supporting certain aspects of the metropolitan planning decision-making process.

Case Study

This research uses a descriptive case study to illustrate the role that a simpler ITLUM can play in the metropolitan planning process. While no two regions are exactly the same, the results of this case study (examining the Washington, DC region) can be used to generalize how to deploy a simpler ITLUM tool since detailed data are identified regarding model requirements, calibration, and testing[footnoteRef:4]. The data available for the Washington, DC region is similar to other regions in the U.S. and many regions are facing similar challenges regarding transportation and land use policies. [4: See Hardy (2011): Simplified Integrated Transportation and Land Use Modeling to Support Metropolitan Planning Decisions: An Application and Assessment for a complete documentation of data sources and calaculation.]

The case study method has been described by some researchers as a weak research method within the realm of social science. Critics point to the insufficient precision of the results, lack of objectivity by the researcher, lack of generalizability, and the limited academic rigor. Nonetheless, proponents point to recent evidence suggesting that it is a commonly used method that, if conducted properly, can mitigate concerns raised by critics. There are three key criteria that can be used to judge the quality of descriptive case studies [15]. First is “construct validity” or the establishment of the correct operational measures for the concepts being studied. In this case study, the operational measures focus on how the Washington Region model is constructed and then how it could be used. Second is “external validity” or the ability to make generalizations from the case study’s findings. While every region in the U.S. is unique in some way, including the Washington, DC region, this region is typical of many regions in the U.S. in terms of data availability and transportation/land use policies being debated. Thus, the findings are applicable to other regions in the U.S. (though not necessarily other countries). Third is “reliability” or the ability to repeat the operations of the study with the same results. An important component of developing this case study is documenting exactly where the data comes from and how the simpler modeling approach is created.

By addressing the issues raised by critics, the case study method can be used as an effective investigative tool. The case study approach is useful when addressing research questions associated with “how” something can be used; where the researcher has little control over events; and the focus is on contemporary phenomena within a real-world context [15]. In other words, the case study approach is valuable when the context of the problem being explored is an important component of the analysis and comparative data is not available due to the contemporary nature of the problem at hand.

MARS Model

This research employs the Metropolitan Activity Relocation Simulator (MARS) model as a simplified ITLUM tool[footnoteRef:5]. The MARS model is a system dynamics model originally developed in Vienna, Austria by Pfaffenbichler [16]. Subsequently, it has been applied in sixteen European and Asian cities and one South American city (Porto Alegre, Brazil)[footnoteRef:6]. The MARS model is deterministic in nature, meaning there are no stochastic elements built into the model. While a stochastic model would yield different results each time it is run, the MARS model will yield the same results each time it is run unless an exogenous variable is changed prior to the start of the model run. The deterministic nature of MARS is one characteristic of it being a simpler tool. This research is the first application of the MARS model in the U.S. and is based upon the application of the MARS model to the city of Leeds, England. [5: A complete description of the MARS model is available in Pfaffenbichler (2008). What is included in this section is a qualitative description of the MARS model structure and functionality.] [6: For a detailed description of where the MARS model has been applied, see Pfaffenbichler (2008).]

The MARS model is an ITLUM tool consisting of two basic sub-models: the transportation model and the land use model. These two sub-models represent both the demand (land use) and supply (transportation) of a metropolitan region. Changes in the transport system cause time- lagged changes in the land use system and changes in the land use system cause immediate reactions in the transport system. The land use sub-model can be further subdivided into a residential and a workplace location sub-model. The links between the sub-models are shown in Figure 1.

Network Development

The purpose of the description of the network development is to indicate the tractability of obtaining the necessary data for the Washington DC MARS Model. Obtaining the required data to support a transportation and land use model is often seen as one of the largest barriers to their use [8]. Thus, developing a transportation and land use model that does not require the collection of unique and specific data for a specific region, but one that can use readily-available data could be seen as an important aspect of using the model. To that end, open sources of data were used to populate the Washington DC MARS Model and the following data sources were used:

· City-data.com

· GMU Center for Regional Analysis

· Google Maps

· Google Transit Feed Specification (GTFS)

· National Household Travel Survey (NHTS)

· WashCOG Round 7.1 Cooperative Forecast

· U.S. Census Bureau

· U.S. Geological Survey (USGS) Land Cover Data Set

The first step in developing a MARS model is to identify individual MARS analysis zones that are consistent in terms of land use type and density which is the similar process used to develop traffic analysis zones in four step models (Pfaffenbichler, Günter Emberger, and Shepherd 2008). The Washington DC MARS Model network was developed based upon existing research conducted by the Washington Council of Governments (WashCOG). The 2,191 traffic analysis zones and 59 regional activity centers and clusters were used to identify 97 individual MARS analysis zones.The second step in developing a MARS model is to collect the necessary data describing the individual MARS analysis zones as well as the travel characteristics, in aggregate, among the analysis zones. Four categories of data are required: regional data, zonal data, passenger car data, and public transportation data. A detailed description of developing the required input data is available in Hardy 2011.

A major hurdle in using any transportation-based simulation model is the development of the transportation network. When first examining the use of the MARS model for this research, developing the transportation network (both roadways and transit) appeared to be the most daunting task even with the simplification of the transportation network to a single link between each origin/destination pair. However, the use of web-based tools such as Google Maps and the GTFS significantly improved this aspect of building the model. In fact, since Google Maps covers all of the U.S., using the procedures documented in this research could be easily replicated in other regions. Also, GTFS continues to evolve with many transit systems publishing schedule data in the GTFS format.

Calibration

Model calibration and validation are important considerations of any simulation model, be it a sketch planning tool such as the MARS model or an agent-based microsimulation model. For purposes of this research, model calibration is defined as the process of estimating the model inputs and parameters such that the output of the model fits an observed data set. The process of model validation uses a calibrated model and compares model outputs with a secondary observed data set. Model validation of the Washington Region MARS Model is undertaken in the following section, Reasonableness Checking.

The MARS model requires the calibration of the two sub-models: the transportation sub-model and the land-use sub-model. In keeping with the nature of this research, readily-accessible calibration data sets are used such that the approach of this research can be replicated in other regions of the U.S. For the transportation sub-model, the 2001 National Household Travel Survey (NHTS) data were used. For the land use sub-model, the Census Transportation Planning Products (CTPP) 3-year tabulations were used. Both of these data sets are readily available and cover the entire U.S. The following sections discuss the process that was used to calibrate the transportation and land use sub-models for the Washington Region MARS Model.

Transportation Sub-model

Calibration of the Washington Region MARS transportation sub-model follows the same method as developed by Pfaffenbichler for the Vienna, Austria MARS model [16]. First, total trips are examined by purpose in terms of commuting trips (Home-Work-Home) and other trips (Home-Other-Home). Second, mode splits (total) are examined in terms of car and rail utilization. Finally, trip generation by individual MARS analysis zone is examined. The method developed by Pfaffenbichler is robust in nature and allows the analyst to adjust several different input variables to adjust the model to reflect that of the calibrated data set. In addition, a number of different parameter values can be further adjusted to better reflect real-world conditions. For purposes of this research, parameters developed for the Leeds, England MARS model were initially used and later adjusted in order to establish a calibrated Washington Region MARS model.

The observed data set which was used to calibrate the transportation sub-model is the 2001 NHTS. The NHTS data set provides data on personal travel behavior, trends in travel over time, and trip generation rates to use as a benchmark in reviewing local data, and data for various other planning and modeling applications [17]. The 2001 NHTS data set was used because it provided the necessary data at the required spatial scale (trip generation rates at the traffic analysis zone level). Future analysts will be able to use the 2009 NHTS data set as it is developed over the upcoming years.

The process of developing the necessary calibration data set required manipulation of the 2001 NHTS data set to create a 2005 NHTS Estimated data set. The 2005 NHTS Estimated data set was used to compare to the outputs of the Washington Region MARS model. Results of calibrating the total trips of the transportation sub-model are shown in Table 1. Results of calibrating the mode splits are showing in Table 2. Finally, results of calibrating trip types by zone were also conducted using regression analysis but are not shown here. As seen in the tables, the MARS model was calibrated successfully. The only discrepancy is the overestimation of rail trips by 18.2 percent. The reason for the discrepancy is not clear since a number of different parameters were adjusted with similar results. One conclusion is that the MARS model may not have enough sensitivity to travel time, and selecting one mode over another is not as simple as which is the least cost, but which is more convenient or comfortable, a factor difficult to represent in a simulation model.

Calibration of the Washington Region MARS land use sub-model follows a similar method as developed by Pfaffenbichler for the Vienna, Austria MARS model [16]. The number of residents is calibrated followed by the number of workers. The observed data set which was used to calibrate the land use sub-model is the CTPP 3-year tabulations. The method used in this research is a modified approach since the observed data set is not available at the desirable spatial scale. In order to use the CTPP 3-year tabulation data set to calibrate the Washington Region MARS model, the land use sub-model had to be calibrated at a larger spatial scale than desirable. Due to constraints associated with the analysis of the data, the 3-year tabulations are only available for a geographic region with a minimum population of 60,000 people. For example, the City of Falls Church, VA, has a population of roughly 44,000 people. Thus, the CTPP 3-year tabulation is not available for the City of Falls Church. For this research, the land use sub-model was calibrated to the following eight geographic regions: District of Columbia, Montgomery County (Maryland), Prince George’s County (Maryland), Arlington County (Virginia), Fairfax County (Virginia), Prince William County (Virginia), and Alexandria City (Virginia).

Results of calibrating the residents is shown in Table 3 with the workers shown in Table 4. Overall, the calibration of the land use sub-model yielded acceptable results. In the future, more spatially disaggregate data will likely be available as the CTPP 5-year tabulations are created. However, these data will not likely yield better results. In fact, there will likely be more variability in the difference between the observed and model outputs.

Table 4 Calibration—Workers (All)

.

An important aspect of developing and deploying a computer modeling tool for transportation and land use planning is to validate the model and ensure that the results of the model in terms of its analytical capability are reasonable. One method to do this, and ensure the reasonableness of its outputs and functionality, is to compare the results of a calibrated model against those of an existing model that has been used to conduct similar forecasts. This is the approach undertaken here and uses a study published by the National Capital Region Transportation Planning Board (TPB) examining alternative land use and transportation scenarios as the existing (or baseline) model results [18]. Using the TPB report as the baseline, this research compares the results of the Washington Region MARS model with that of the TPB report in terms of three different measures of effectiveness: land use (residents and workers), mode split, and vehicle travel. The purpose of this analysis is not to determine if the results match absolutely, but whether there are similarities in terms of directionality (e.g., increase or decrease in residences), order of magnitude changes (e.g., 1 percent versus 10 percent change in mode split), and spatial location of the changes in the region. For purposes of this research two of the five scenarios from the TPB report were examined. In addition, a third scenario was created related to road user charges. These three scenarios are described below:

1. Transit Oriented Development for Rail (TOD-Rail)—The Transit Oriented Develop for Rail scenario is designed to test the effects of concentrating more of the region’s growth in areas that could be served by rail transit.

2. Region Undivided (RU)—The Region Undivided scenario is designed to test the effects of enabling workers to live closer to their jobs by assuming shifts in future job and household growth from the western portion of the region (Montgomery County, Maryland; Loudoun County and Fairfax County, Virginia) to the eastern portion of the region (Prince George’s County, Maryland).

3. Road User Charges (RUC)—The Road User Charge scenario is not one of the five included in the TPB report. This scenario is designed to test the effects of increasing road user charges through a RUC fee for the entire region. Low and high scenarios were analyzed.

Model Functionality

First, the three scenarios are compared to the baseline to ensure the results of the are as expected. The three scenarios described above, along with a baseline scenario (e.g., calibrated model), were analyzed using the calibrated Washington Region MARS model. The comparison of the results is for the 2025 forecast year which is consistent with the analysis of the TPB report. The results of the analysis are shown in Table 5. In the table, the values for each measure are shown in the top row with the percent difference calculated for each measure and scenario against the baseline scenario in the bottom row.

The results of the TOD-Rail scenario are generally what one would expect to see. In this scenario, land use development potential was increased in the analysis zones with rail stations. Overall, VMT is projected to decrease while the total number of trips will increase. Within the trips, the number of car trips will be reduced by 3.8 percent and the rail transit trips will increase by 16.7 percent. The other important component of this scenario is the location of where new development for houses and jobs is going to occur. Because this scenario favors development around transit stations, one would expect to see an increase in residences and employment in analysis zones with rail transit stations.

The Region Undivided scenario had different results than the TOD-Rail scenario. In the Region Undivided scenario, the projected increase in workers and residents was forecast to occur in the eastern portion of the region, primarily Prince George’s County, Maryland. Overall, VMT and trips were projected to increase by 2.9 percent and 3.6 percent, respectively. A more detailed analysis of the Region Undivided scenario projected a 2.7 percent increase in both residents and workers in the region with a significantly higher amount (27 percent) occurring in Prince George’s County.

The Road User Charge scenario analyzed the effect that a simple road user fee would have on the transportation and land use system. In the MARS model, the road user fee was modeled as a single additional cost per car trip. The RUC-Low was $5 per trip and the RUC-High was $10 per trip. The results are what one would expect. Total VMT is projected to decrease for both the RUC-Low and RUC-High scenarios. In terms of trips, the total number of trips is projected to decrease, but the number of rail transit trips will increase due to the implementation of the road user charge. Regarding the sensitivity of the Washington Region MARS model to a road user charge, doubling the road user charge yields a 1.8 percent decrease in VMT, a 5.9 percent decrease in car trips, and a 2.9 percent increase in transit trips.

Model Validation

Second, a model validation exercise was undertaken. Given that the results of the model among the different scenarios were as expected and reasonable, the calibrated model was compared to the TPB study results. Table 6 provides a summary of the comparison between the results of the TPB study and the Washington Region MARS model. In both cases the comparison is between the final forecasted year for each model (2030 for the TPB study and 2025 for the Washington Region MARS model). In the table, a shaded cell indicates a discrepancy between the directionality of the measure while underlining indicates a difference in the order of magnitude of the measured values. In general, the results of the Washington Region MARS model track well with the TPB study regarding the directionality of the values.

In all but one instance (VMT measure for the Region Undivided scenario), the directions of the measured change are consistent. For both scenarios, the TPB study projected a decrease in VMT and an increase in mode split and land use. MARS estimated an increase in VMT for the Region Undivided scenario. In examining the land use changes of where residents and workers were locating, it was apparent that there were increases in the outer regions (Prince William and Loudoun County) resulting in additional VMT. It is likely that cost of land in the outer regions is less expensive and is a larger determinant of where new residents and workers choose to locate over the cost of transportation. Thus, there is no indication that because the two model results are not similar that one is either correct or incorrect. A more detailed analysis of the manner in which the TPB study model allocates land and distributes new development would need to be conducted to determine whether the process of land-use allocation is similar in both models.

Where there were significant differences was concerning the order of magnitude of the changes. For example, in the TOD-Rail scenario, the TPB study projected a 0.8 percent decrease in VMT with the Washington Region MARS model projecting a 5.7 percent decrease. At this point, there is no clear indication of why this is occurring. One possible explanation is the manner in which the Washington Region MARS model accounts for transit trips and the lack of sensitivity to the qualitative characteristics of the transit system (see previous discussion on model calibration). However, there is a likely explanation for the differences in the Region Undivided scenario due to the lack of bus transit system representation in the Washington Region MARS model. The TPB study included bus transit systems in the model and also included an increase in bus transit systems when running Region Undivided scenario. Thus, it makes sense that the TPB scenario would result in a larger increase in transit trips.

The central question of concern in this research is what are the opportunities and limitations of using readily-available data sources to populate, calibrate, and validate a simplified ITLUM tool and what are the implications? This research employed a case study approach to assess the tractability of using open sources of data to populate, calibrate, and validate a simplified ITLUM tool. While this research was part of a larger effort, the research summarized here demonstrated the ability to a) calibrate a simplified ITLUM tool using readily-available data and b) validate the results of the calibrated model. The ability to develop a simplified ITLUM tool using open data sources should be of concern to those planning organizations facing the need to analyze many different scenarios. The use of traditional travel demand models may not lend themselves to this task, but the use of simplified tools, such as the ITLUM tool used is this research, could be uniquely qualified.

The implications of this research are important. This research demonstrates a new evolution to the development of ITLUM tools. Historically, model development has been described as a process of evolving tools from simple to complex with the goal of developing tools that are, generally speaking, more complex in nature [3], [4]. The concern now is not on model complexity (can I get a better answer with a more complex tool?) but rather the requirements of the model in terms of data, resources (time and money), functionality, and expertise and the decision-making function that a computer modeling tool supports. Future research will need to examine not just the complexities of the model, but what types of decisions the model is supporting.

REFERENCES

[1] K. A. Abbas and M. G. H. Bell, “System dynamics applicability to transportation modeling,” Transportation Research Part A: Policy and Practice, vol. 28, no. 5, pp. 373-390, Sep. 1994.

[2] A. Mostashari and J. Sussman, “Stakeholder-Assisted Modeling and Policy Design Process for Environmental Decision-Making,” Journal of Environmental Assessment Policy & Management, vol. 7, no. 3, pp. 355-386, 2005.

[3] E. Miller, D. Kriger, and J. D. Hunt, Integrated Urban Models for Simulation of Transit and Land Use Policies: Guidelines for Implementation and Use. University of Toronto Joint Program in Transportation, 1998.

[4] M. Wegener, “Overview of Land Use Transport Models,” in Handbook of Transport Geography and Spatial Systems, D. A. Hensher and K. J. Button, Eds. Amsterdam: Elsevier, 2004, p. 672.

[5] D. Mile and G. Emberger, “Providing for Mobility: Transportation Planning Under Pressure,” in Twenty-First Century Leeds: Geographies of a Regional City, Revised edition., PLACE Research Centre, York St John College, 2004.

[6] M. Hardy, J. Larkin, K. Wunderlich, and A. Nedzesky, “Estimating User Costs and Economic Impacts of Roadway Construction in Six Federal Lands Projects,” Transportation Research Record: Journal of the Transportation Research Board, vol. 1997, no. 1, pp. 48-55, Jan. 2007.

[7] M. Hardy, K. Wunderlich, and J. Bunch, Structuring Modeling and Simulation Analyses for Evacuation Planning and Operations. Washington, DC: U.S. Department of Transportation, 2009.

[8] U.S. Government Accountability Office, Metropolitan Planning Organizations: Options Exist to Enhance Transportation Planning Capacity and Federal Oversight. 2009.

[9] M. Iacono, D. Levinson, and A. El-Geneidy, “Models of Transportation and Land Use Change: A Guide to the Territory,” Journal of Planning Literature, vol. 22, no. 4, pp. 323-340, May. 2008.

[10] C. Porter, L. Melendy, and E. Deakin, Land Use and Travel Survey Data: A Survey of the Metropolitan Planning Organizations of the 35 Largest U.S. Metropolitan Areas. U.S. Department of Transportation, 1995.

[11] J. Sussman, S. Sgouridis, and J. Ward, “New Approach to Transportation Planning for the 21st Century: Regional Strategic Transportation Planning as a Complex Large-Scale Integrated Open System,” Transportation Research Record: Journal of the Transportation Research Board, vol. 1931, no. 1, pp. 89-98, Jan. 2005.

[12] D. A. Badoe and E. J. Miller, “Transportation-land-use interaction: empirical findings in North America, and their implications for modeling,” Transportation Research Part D: Transport and Environment, vol. 5, no. 4, pp. 235-263, Jul. 2000.

[13] P. Waddell, G. F. Ulfarsson, and J. P. Franklin, “Incorporating land use in metropolitan transportation planning,” Transportation Research Part A: Policy and Practice, vol. 41, no. 5, pp. 382-410, Jun. 2007.

[14] Fehr & Peers, An Assessment of Integrated Land Use/Transportation Models. Southern California Association of Governments, 2007.

[15] R. K. Yin, Case Study Research. SAGE, 2008.

[16] P. Pfaffenbichler, MARS - Metropolitan Activity Relocation Simulator. VDM Verlag Dr. Mueller e.K., 2008.

[17] FHWA, “2001 National Household Travel Survey User’s Guide Version 3.” Federal Highway Administration, Jan-2004.

[18] D. McAuslan and K. Ransome, Regional Mobility and Accessibility Study: Alternative Land Use and Transportation Scenarios. Washington, DC: National Capital Region Transportation Planning Board, 2006.

16