Office of Planning, Environment, & Realty (HEP)
Planning · Environment · Real Estate
A multiyear project at the University of California-Davis has compared the use of a regional travel demand model, SACMET96, with two transportation-land use models, MEPLAN and TRANUS, for testing regional transportation and land use policies. The project has evaluated a range of policies, both individually and in combination, including:
High-occupancy vehicle (HOV) lanes;
High-occupancy toll (HOT) lanes;
Light rail transit (LRT) and other advanced transit;
Transit-Oriented Development (TOD);
Roadway-Oriented Development (ROD); and
Pricing.
Impacts were measured for the years 2005 and 2015 on travel, emissions, user benefits, and the spatial distribution of population and employment. The results of the modeling are not always intuitive. Some of the major findings for the Sacramento region include:
Accounting for land use effects can have significant impacts on forecast vehicle-trips, VMT, congestion, and emissions. Travel and emissions impacts were found, in general, to be significantly greater in MEPLAN than in SACMET96 for comparable policy scenarios.
Some policies are synergistic while others may work at odds. For example, transit and pricing policies had little impact individually but a more significant impact in combination. While parking pricing had the effect of reducing trip lengths and VMT, it also provided a disincentive to transit station area development and thus had little impact on mode share.
Similar to transit investments, HOV and HOT policies were found to be much more effective when applied in conjunction with land use and/or pricing policies.
Positive economic user benefits were found for most scenarios. When applied by income group, the user benefit measure could be used to assess the equity implications of each scenario.
The Sacramento metropolitan area is a mid-size urban area with an estimated 1995 population of 1.8 million. Population and employment are expected to grow annually at rates of 1.9 and 2.2 percent, respectively, through 2015. In the past, the employment base of the Sacramento region has been largely government and agriculture; however, more recently there has been a rapid expansion of high-technology manufacturing. The residential and employment densities of the region can be characterized as medium to low. Current mode shares for home-based work trips are approximately 76 percent drive alone, 17 percent carpool, three percent transit, two percent walk, and two percent bike.
Figure 1 shows a map of the Sacramento region and major transportation facilities. Interstate 80 and U.S. Highway 50 are major east-west routes through the area; Interstate 5 and U.S. 99 run north-south. Regional growth has largely been occurring to the northeast and east, into the foothills of the Sierra Nevada, and to the south in the Sacramento Valley. A light rail system was first opened in 1987 and currently totals 20.6 miles in length. Two more corridors adding 17.2 miles are slated to open by 2003.
Source: Johnston , Rodier, Choy, and Abraham (2000).
The work documented in this case study is part of a multiyear project directed by Professor Robert Johnston at the University of California at Davis. The purposes of the project are: 1) to compare the utility of different transportation and transport-land use models at regional policy testing; 2) to test alternative transportation and land use policy scenarios in the Sacramento region; and 3) to demonstrate methods that metropolitan planning organizations (MPOs) can apply to test similar scenarios in other regions. The researchers have worked with the Sacramento Council of Governments (SACOG) to obtain data and models and to formulate realistic transportation and land use alternatives.
The results have not been applied specifically in developing the regional transportation plan or regional land use policies. However, a review of SACOG's Environmental Impact Review of the Metropolitan Transportation Plan (MTP) suggests the potential for future applications. The EIR references a number of issues - such as land use distributional impacts - that it would be "speculative to measure at this time." (p.24) The development and application of modeling techniques such as those described here could facilitate a broader understanding of the various environmental impacts considered in the planning process.
The Sacramento analysis is the first application of MEPLAN in the U.S., but it is not the only application of a transportation-land use model. A number of metropolitan areas have used the Disagreggated Residential Allocation Model/Employment Allocation Model (DRAM/EMPAL) to forecast land use as a function of transportation accessibility (Putman, 1996). DRAM/EMPAL, however, is limited in its ability to test the impacts of land use-related policies. The TRANUS modeling system (De la Barra, 1989) is closely related to MEPLAN, and is currently being applied in Baltimore, MD to test the impacts of "Smart Growth" policies (see Box 1). UrbanSim is a model developed in the U.S. by Paul Waddell of the University of Washington under the National Cooperative Highway Research Program (NCHRP). UrbanSim has been applied in Honolulu, Portland, and Salt Lake City. The model contains a number of promising advances that should improve the quality of urban modeling further and will continue to expand the range of policies that can be modeled.
In a related analysis for the Sacramento region, Johnston and de la Barra (2000) show how land use projections from an economically based urban model can be disaggregated to small areas using GIS (see Box 2). The results in turn can be used to estimate the footprint of urban development, and to derive estimates of runoff, sedimentation, habitat impacts, and development-related costs.
Box 1. Applying TRANUS to Model "Smart Growth" in Baltimore
The Baltimore Metropolitan Council is in the process of calibrating TRANUS, an integrated transportation-land use model, for the region. The model will be used to evaluate the impacts of transportation investments on land development, as well as the transportation, economic, and environmental impacts of "smart growth" policies. The model will also be used to evaluate impacts on different income groups, using consumer surplus measures.
Contact the Baltimore Metropolitan Council for more information, or see Liu (1998) for a description of the model structure.
Box 2. Linking Urban Models and GIS
Johnston and de la Barra (2000) demonstrate the benefits of linking a regional land use-transportation model with a GIS to identify physical and environmental impacts of development. Their application is based on the TRANUS model which they developed for the Sacramento area. TRANUS is in many respects similar to the MEPLAN model, which was also applied in Sacramento as described in this case study.
The Sacramento TRANUS application includes 58 internal districts. Forecasts for each district were developed by land use category, including agricultural/mining/forestry, industrial, office high density, office low density, residential high density, and residential low density.
Once district-level land use forecasts were produced, these forecasts were disaggregated to small areas using the California Urban Futures Model (CUFM), developed by John Landis at the University of California - Berkeley (Landis, 1995). CUFM allocates development to polygons, which are ranked according to profitability for the developer. Profitability can be determined by a variety of variables such as local government fees and land prices; in this analysis it is calculated based on accessibility. Developable land units are created by overlaying GIS coverages such as city boundaries, wetlands, slope, land use type, and roads. A total of 272,000 polygons were created in this application.
A potentially quicker alternative to polygon-based disaggregation would be the use of a cell-based (raster) mapping technique, as described in the Envision Utah and SPARTACUS case studies. Using either technique, the amount of projected development by type can then be overlaid with other GIS layers, such as sensitive habitats or agricultural land. Functions can also be applied to estimate impermeable surface area, infrastructure costs, or other impacts according to the type and density of development.
Standard travel demand models are capable of measuring the transportation impacts of alternative land development patterns. There are at least three reasons, however, to apply a transportation-land use model such as MEPLAN, TRANUS, or UrbanSim (Figure 2):
To measure the effect of alternative transportation investments and policies on land use patterns (1);
To consider the resulting feedback of these land use changes on transportation performance and related impacts (2); and
To measure the effect of alternative land use policies, such as zoning or tax incentives, on land use patterns and the resulting transportation impacts (3).
Figure 2. Relationships Modeled by Travel Demand and Land Use Models
In this study, two separate modeling approaches are used and the results are compared. The approaches are:
Using a travel demand model only. Under this approach, alternative transportation investments and policies are simulated by changing network characteristics in the travel demand model. The effects of land use policies are simulated by changing the geographic distribution of future development. This approach is used widely by metropolitan areas throughout the United States.
Using a travel demand-land use model. Under this approach, an integrated model is used that includes feedback between transportation accessibility and land development. It also allows land use policies, such as development fees or zoning changes, to be modeled directly by influencing the price and availability of land.
Under both approaches, changes in emissions are then calculated from transportation system changes using an emission factor model.
This study uses the 1996 Sacramento regional travel demand model (SACMET96). According to the study authors, the SACMET96 model includes a number of features that make it well-suited for modeling land use, transit, and pricing measures. Some of these features include:
Feedback of travel impedances to the trip distribution step;
Accessibility as a determinant of auto ownership and trip generation;
A joint destination and mode choice model for work trips;
A mode choice model with separate walk and bike modes, walk and drive transit access modes, and two carpool modes (two and three or more occupants);
Land use, travel time and monetary costs, and household attribute variables included in the mode choice models;
All mode choice equations in logit form; and
A trip assignment step that includes separate a.m., p.m., and off-peak periods.
The mechanism for incorporating land use into the mode choice models is through a pedestrian environment factor (PEF). The PEF is a rating assigned to each zone based on how conducive that zone is to pedestrian travel. The rating is similar to that used in the Portland LUTRAQ study and is based on sidewalk availability, ease of street crossing, connectivity of the street/sidewalk system, and terrain. Additional documentation of the SACMET96 model can be found in DKS (1994).
The study authors also note that the SACMET96 model, like many travel demand models, contains a number of limitations that mean that not all transportation choices are fully sensitive to the range of policy options tested. For example:
In the trip distribution step, non-work trips are sensitive only to travel time and not to travel cost;
Time-of-day factors are applied after mode choice and, as a result, only the work trip purposes use peak or congested travel times during the trip distribution and mode choice steps;
The trip assignment step is not directly sensitive to travel cost and only the home-based work trips use congested times. As a result, the models sensitivity to the HOT lanes and pricing policies is somewhat limited;
Only the home-based work mode choice model uses three income groups; other mode choice models are not based on income groups, which restricts the sensitivity of the simulation;
SACMET96 does not include a time-of-day choice model and cannot simulate the phenomenon known as peak spreading.
The Travel Model Improvement Program has identified short-term and long-term ways to enhance travel demand models to more accurately model a broader range of transportation investments and policies.
MEPLAN
The MEPLAN framework draws on over 25 years of spatial economic modeling experience and has been used around the world in cities such as Naples, Helsinki, and Bilbao, but the Sacramento model is the first application in the U.S. Moreover, this is the first study in which an integrated land use and transportation model uses separate a.m., p.m., and off-peak assignment models (as opposed to an average daily assignment model) for more accurate emissions analysis.
Abraham and Hunt (1999) provide a description of the MEPLAN model, from which the following description is drawn. The basis of the MEPLAN modeling framework is the interaction between two parallel markets, the land market and the transportation market. This interaction is illustrated in Figure 3. Behavior in these two markets is a response to price signals that arise from market mechanisms. In the land markets, price and generalized cost (disutility) affect production, consumption, and location decisions by activities. In the transportation markets, money and time costs of travel affect both mode and route selection decisions.
The cornerstone of the land market model is a social accounting matrix (SAM). The SAM is an input-output table that is expanded to include relationships among four factors: industries, households, building floor space, and land. The SAM will tell, for example, that one unit of manufacturing industry requires X units of floor space, Y units of industrial land, and Z workers, as well as specified amounts of inputs from other economic sectors. (Input-output tables can be obtained from the Bureau of Economic Analysis' Regional Input-Output Modeling System or from IMPLAN.) The SAM is spatially disaggregated at the zonal level. Logit models of location choice are used to allocate volumes of activities in the different sectors of the SAM to geographic zones. The attractiveness or utility of zones is based on the costs of inputs to the producing activity (which include transportation and rents), location-specific effects, and the costs of transporting production to consumption activities.
The resulting patterns of economic interactions among activities in different zones are used to generate origin-destination matrices of different types of trips. These matrices are loaded to a multimodal network representation that includes nested logit forms for the mode choice models and stochastic user equilibrium for the traffic assignment model (with capacity restraint). The resulting network times and costs affect transportation costs, which then affect the attractiveness of zones and the location of activities, and thus the feedback from transportation to land use is accomplished.
Figure 3.
The Interaction of the Land Use and Transportation Markets in the MEPLAN Framework
Source: Johnston, Rodier, Choy, and Abraham (2000).
The framework is moved through time in steps from one time period to the next, making it "quasi-dynamic." In a given time period, the land market model is run first, followed by the transportation market model, and then an incremental model simulates changes in the next time period. The transportation costs arising in one period are fed into the land market model in the next time period, thereby introducing lags in the location response to transport conditions. See Hunt (1994), Hunt and Echenique (1993) or Abraham and Hunt (1999) for descriptions of the mathematical forms used in MEPLAN.
Sacramento MEPLAN Model
As described in Johnston, Rodier, Choy, and Abraham (2000), the Sacramento MEPLAN model uses 11 industry and service categories that are based on the SAM and aggregated to match available employment and location data. Households are divided into three income categories (high, medium, and low). There are seven land use categories in the model. Constraints are placed on the amount of manufacturing land use to represent zoning regulations that restrict the location of heavy industry. Each of these land uses (except agricultural) locates on developed land represented by the factor "urban land." Two factors are used to keep track of the amount of vacant land available for different purposes in future time periods, and the development process converts these two factors to "urban land." A calibration parameter allows differential rents to be paid by different users of the same category of land.
In most applications, including Sacramento, MEPLAN is a "sketch-planning" tool in the sense that the zone system and transportation network are quite coarse. In the Sacramento application, there are 32 analysis zones. The level of detail is primarily a result of the difficulty both in obtaining the required data and calibrating the model for a more detailed zone structure. An automated calibration procedure recently developed for the MEPLAN model (Abraham and Hunt, 2000) may help overcome this limitation to some extent. Another option would be to use MEPLAN and the regional travel demand model in conjunction, by taking land use results from MEPLAN and adjusting land use inputs into the travel demand model accordingly. This approach is being taken by the Baltimore Metropolitan Council, which is in the process of calibrating TRANUS - a land use model similar to MEPLAN - for the Baltimore region (see Box 1).
Data Requirements
MEPLAN can be developed for an area using locally available data sources. The following data were used in developing the Sacramento model:
Households by income and zone for 1990;
Employment by industry and zone for 1990;
Supply of zoned land by zone for 1990 (these data were previously developed by SACOG, in preparation for implementation of the DRAM/EMPAL model);
Average prices for zoned land by category and zone for 1990 (obtained from the TRW-REDI datasets on real property sales by county, a private, nationwide source);
Social accounting matrices for 1985 and 1990;
Transportation networks by mode for 1991;
Trips aggregated into various categories of purpose, mode, and time of day for 1991;
Distributions of travel distances by purpose for 1991, from the SACMET travel demand model; and
Origin-destination matrix of total trips for 1991, based on a sample of households.
According to the study authors, the most important data limitation was a lack of information on building stock, specifically, on the amount of floorspace by building type in each zone. As a result, floorspace was not included directly in the model. Instead "space" variables were developed that approximately represented the amount of developed land.
A related problem was a lack of data to calibrate models of developer behavior. Data on floorspace through time would have facilitated the development of such a model. Without these data, theoretical constructs and rules-of-thumb were used to predict the location and extent of development.
Data on freight flows in the Sacramento region and data on the relationship between economic links and trip rates for freight volumes, were also desired in order to model benefits to goods movement. Unfortunately, these data were lacking, so a "goods movement" flow was omitted from the Sacramento version of MEPLAN. For an example of data collection on commodity flows that would assist in freight model development, see the Portland case study.
The study authors report that the Sacramento MEPLAN model was designed around the available data, and unanticipated trends and inconsistencies in the data led to changes in the model design. Some of these changes were improvements, others might be considered compromises. The project demonstrates that a useful model of urban land use and transport interaction can be developed within this framework in the United States using existing data sources.
Calibration
The parameters in the MEPLAN land use and transport interaction model of the Sacramento, California region were estimated using the sequential approach. (Calibration of the Sacramento model is discussed in Abraham and Hunt, 1998; the sequential approach and alternative calibration approaches are discussed in detail in Abraham and Hunt, 2000. The authors have developed an automated calibration approach that should expedite model application in the future.) Time-series data would have been preferable, but the available data were sufficient to obtain a working model. The following sets of measured variables were used to formulate the objective function for estimating the overall parameters:
Various indications of the proportion of travel occurring by different modes;
The spatial distribution of trips in the form of two origin-destination matrices, one for private automobile trips and another for trips by other modes; and
The "trip length distributions" showing, for automobile trips, the proportion of trips that are of different lengths.
The California Air Resources Board's EMFAC7F model was used to develop vehicle emission factors, and The California Department of Transportation's Direct Travel Impact Model 2 (DTIM2) was used to simulate the emissions effects of scenarios. Travel model outputs from the a.m. peak, p.m. peak, and off-peak periods were used in conjunction with regional cold start and hot start coefficients for each hour in a 24-hour summer day, as provided by SACOG.
For metropolitan areas outside California performing a similar analysis, emission factors from the EPA's MOBILE model could be used. The MOBILE6 model, to be released in 2000, will provide the advantage over MOBILE5 that it will explicitly model emissions from starts as well as VMT (the California models already have this capability). This can be important because different scenarios may have different effects on the number of trips vs. trip lengths. A transit scenario without transit-oriented development, for example, may cause a larger decrease in VMT than trips, since many people will access the transit system via a car trip.
In addition to travel, emissions, and land use changes, the study also measured overall user benefits. The total monetary value of cost savings, time savings, and other benefits were measured based on the difference in the overall "utility" of travel between two alternatives. Capital and operating costs for each alternative are then subtracted, to obtain the net change in benefits. User benefits are calculated for both personal and commercial vehicle travel. Overall benefits are reported on a cost-per-trip basis.
The method for calculating compensating variation is described in Box 3.
Box 3. Calculation of User Benefits Using Compensating Variation
As discussed by Johnston, Rodier, Choy, and Abraham (2000), Small and Rosen (1981) show how a consumer benefit measure known as compensating variation (CV) can be obtained from discrete choice models:
where l is the individual's marginal utility of income, Vm is the individual's indirect utility of all m choices, p0 indicates the initial point (before the policy change), and pf indicates the final point (after the policy change). The change in indirect utility is converted to dollars by the factor, 1/l, or the inverse of the individual's marginal utility of income.
The compensating variation formula (1) above is adapted to work with the SACMET96 mode choice models. In these models, households are segmented into income/worker categories and person trips are generated for those categories. To obtain compensating variation for each income/worker category h, the following formula is applied for all modes m and for all trips Q between all origins i and all destinations j:
where l is provided by the coefficient of the cost variable in the mode choice equations. Total compensating variation is obtained by summing the compensating variation obtained from each income/worker group. Compensating variation is also obtained from the non-home-based mode choice models, which are not stratified by household/income classes.
Compensating variation can easily be calculated by extracting the required information from the travel model and manipulating it in a spreadsheet. The specific steps are as follows:
Write a program within the travel model (in this case, MinUTP) to output the logsum (denominator) of the mode choice equation for each trip type and household/income category, as well as the number of trips in each category.
Import these data into a spreadsheet, and calculate CV for each category. Summing this value would provide the total user benefits, without including project costs.
Determine the estimated annualized capital and operating costs of each alternative.
Determine the proportion of VMT attributable to each trip category, based on number of trips and average trip length by type.
Allocate the annualized capital and operating costs to trip category based on the proportion of VMT in that trip category.
For each trip category, subtract the allocated costs from the CV to produce the net economic benefit. Sum across trip types to get net benefits by income group.
Based on a review of the literature, the authors assumed total operating costs of $0.40 per mile. Capital and operation and maintenance (O&M) costs of the new facilities were estimated based on cost figures provided in the Sacramento region's 1996 metropolitan transportation plan.
Commercial Vehicle Travel
Economic benefits for commercial vehicles are calculated based simply on travel time and operating cost savings. Benefits are obtained from the trip distribution model in SACMET96, which distributes commercial vehicle trips as a function of zone-to-zone travel times. As reported in Johnston, Rodier, Choy, and Abraham (2000), the following formula was applied:
where B is equal to the net benefits to commercial vehicle travel, including travel time costs, O&M costs, and revenue benefits. Travel time and VMT changes for commercial vehicles are obtained from the travel model for each scenario. Travel time is converted to dollars using the average wage rate of truck drivers in the region ($12 per hour). Total O&M costs and revenue benefits for the scenarios, excluding wages, are obtained by multiplying the average per mile costs for the region ($0.90 for O&M and $0.95 for revenues) by VMT. (Truck wages, O&M costs, and revenues were developed based on national data and in consultation with the California Trucking Association; very little local data were available.)
User benefits were calculated using SACMET96 but not MEPLAN. There are differences of opinion over whether consumer surplus measures, such as the Small-Rosen traveler welfare model described above, are valid when land use demand shifts. Using the Portland statewide model (a TRANUS application), Dr. Jonathan Hunt at the University of Calgary is currently testing methods to obtain measures of consumer welfare from MEPLAN, TRANUS, or a similar transportation-land use model.
For additional discussion of this measure, see Rodier, Johnston, and Shabazian (1998).
The level of effort described here focuses on the application of MEPLAN, since the regional travel model analysis (SACMET96) is a straightforward application of standard MPO travel modeling tools. Total estimated staff time for the Sacramento MEPLAN analysis was seven to 14 person-months (PMs), including two to three person-months of MEPLAN staff time at the University of Calgary. The total cost for the analysis was roughly $50,000 for UC-Davis researchers and $20,000 for MEPLAN staff time. The analysis described here was a "quick and dirty" application; more extensive and thorough data and model development would add an additional level of effort.
The level of effort broken down by task was roughly as follows:
Developing the network and zone structure for MEPLAN: about one PM. This can now be edited in ArcView, although it may be easier to have a person familiar with MEPLAN do it.
Developing baseline data for MEPLAN: The cost depends on the state of the datasets at the MPO. It is helpful to have all of the various data described above, including land use, travel survey, and related data, for two base years. If all these data are available, the effort is a PM or two. The typical effort required in a large metropolitan region might be two to six PMs.
Calibrating MEPLAN: one to two PMs by a MEPLAN staff person.
Developing policy/scenario inputs: relatively easy, except to develop new networks for roads, bus, rail, etc. Complex scenarios might take one to two PMs of MEPLAN staff time, unless the MPO takes the time to learn how to edit the MEPLAN network.
Running MEPLAN scenarios and processing output: one to two PMs for a few scenarios.
The MEPLAN application described here makes use of a high-quality travel model and good travel data. Desirable travel model features include auto ownership feeding elastic trip generation; elastic trip distribution, that is, iterated across all submodels to equilibrium; at least three time periods; a logit mode choice model; and bus, walk/bike, and rail modes, if applicable. If some or all of these features are not included, sensitivity to land use policies and feedback will be reduced. The Sacramento application did not include heavy trucks, however. If the MPO has a truck freight model, an additional two to three PMs would be required, but the ability to track freight movement and also to model commercial land uses would be greatly improved.
Land use data were developed by the researchers. High-quality land sales price data were the most difficult to obtain, and the large zones in MEPLAN also caused problems with averaging over a large ranges of data values. The above level-of-effort estimate is also with only a land model, not a floorspace model. A floorspace model is recommended to provide a check on land consumption and also to represent land use controls more accurately. Floorspace models also track building types by occupying activity, so that new activities must rehabilitate the building. This slows down turnover of activities and is more realistic. If floorspace data and lease payment data (by activity by district) are not readily available, this would add a significant data gathering or estimating effort, adding perhaps six to 12 PMs of effort.
This section describes the specific policy scenarios tested. SACMET96 and MEPLAN are both used to compare various combinations of transit expansion, transit-oriented development (TOD), and pricing scenarios in the years 2005 and 2015. (The land use strategies are examined in 2015 only, because of the long timeframe required to implement such policies and to realize their effects.) All the transportation network improvements are made in the year 2005 for the MEPLAN scenarios, and thus land use is affected in the years 2010 and 2015 because of the five-year iteration period. Because of the different model structures, some of the scenarios had to be modeled differently in SACMET96 and MEPLAN. Therefore, not all of the results are directly comparable.
The project also examined combinations of road-based scenarios, including high-occupancy vehicle (HOV) lanes, high-occupancy toll (HOT) lanes, road-oriented development (ROD), and pricing. These scenarios were modeled using SACMET96 only, so that travel and emission impacts were estimated, but not land use impacts.
Table 1 identifies which scenarios were modeled using which model. Table 2 provides a description of the scenario elements and modeling approaches.
Table 1. Scenarios Modeled
Scenario Name |
SACMET96 |
MEPLAN |
|---|---|---|
Base Case |
X |
X |
Pricing |
X |
|
Transit Scenarios |
||
Light Rail Transit (LRT) |
X |
|
Advanced Transit |
X |
|
Light Rail + Pricing |
X |
|
Transit-Oriented Development (TOD) + Light Rail + Advanced Transit |
X* |
X* |
TOD + Light Rail + Advanced Transit + Pricing |
X* |
|
Road-Based Scenarios |
||
High-Occupancy Vehicle (HOV) Lanes |
X |
X |
High-Occupancy Toll (HOT) Lanes |
X |
|
HOV + Road-Oriented Development (ROD) |
X* |
|
HOT + ROD |
X* |
|
HOV + ROD + Pricing |
X* |
|
HOT + ROD + Pricing |
X* |
|
*Modeled for 2015 only.
This section highlights a few of the findings on individual and combined effects of transit, land use, and pricing scenarios using SACMET96. Scenarios were evaluated with respect to vehicle-trips, vehicle-miles of travel (VMT), vehicle-hours of delay (VHD), mode shares, and VOC and NOx emissions. A much more detailed discussion and interpretation of results is provided in Johnston, Roder, Choy, and Abraham (2000). All comparisons are to the baseline scenario unless noted.
Transit scenarios - In 2005, the effects of the transit-only scenarios are small (roughly a one percent reduction in trips and VMT). The pricing and transit/pricing strategies have larger effects, reducing VMT by four to five percent and VHD by 20 to 25 percent. In 2015, the addition of TOD to the transit scenarios greatly increases their effectiveness at reducing VMT, providing a six percent reduction without pricing and a nine percent reduction with pricing, compared to the baseline scenario (Figure 5).
Highway scenarios - The HOV scenarios exhibit results similar to the HOT scenarios. Both scenarios alone lead to slight increases in trips and VMT and small reductions in vehicle delay. When combined with road-oriented development patterns, VMT is reduced by about two percent and delay is reduced by 11 to 16 percent. When additionally combined with pricing, VMT is reduced by four to five percent and delay is reduced by over 20 percent. None of the HOV or HOT scenarios led to a significant reduction in vehicle-trips (Figure 13).
Emissions - As would be expected, emissions impacts closely followed VMT impacts, although impacts varied somewhat depending on whether vehicle-trips were reduced as well. The authors found that land use intensification policies may be more effective in reducing both VMT and emissions when combined with transit than with HOV- or HOT-lane policies.
User benefits - In the 2005 scenarios, the pricing and transit/pricing scenarios produce the highest economic benefits, eight to nine cents per trip, compared to benefits of one to two cents per trip for the scenarios without pricing. In the 2015 scenarios, the addition of TOD to the transit scenarios significantly increase the economic benefits compared to 2005, with benefits of 10 to 15 cents per trip. In this case, significant user benefits are achieved even without pricing policies. For the road-oriented scenarios, the HOT scenarios provided greater user benefits than the corresponding HOV scenarios.
Interpretation of User Benefits In this method of assessing user benefits, a benefit of 10 cents per trip can be interpreted to mean that the average traveler would be willing to pay an additional 10 cents per trip to have the scenario travel conditions rather than the Base Case travel conditions. |
Equity - User benefits were calculated separately for the three income groups in the model (Figure 7). The results show that light rail alone has a relatively small impact on each income group. Pricing has a positive impact on the middle- and upper-income groups, but a negative impact on the lowest-income group. The transit/TOD scenarios provide positive benefits to all income groups. The addition of transit/TOD to pricing eliminates the regressive effect of pricing by itself, but still provides the greatest benefits to the highest-income group.
Figure 4.
Transit and Land Use Scenarios: SACMET96 Model
Travel, Percent Change from Base Case (2005)
Source: Johnston, Rodier, Choy and Abramham (2000).
Figure 5.
Transit and Land Use Scenarios: SACMET96 Model
Travel, Percent Change from Base Case (2015)
Source: Johnston, Rodier, Choy and Abramham (2000).
Figure 6.
Transit and Land Use Scenarios: SACMET96 Model
NOx Emissions, Percent Change from Base Case
Source: Johnston, Rodier, Choy and Abramham (2000).
Figure 7.
Transit and Land Use Scenarios: SACMET96 Model
Economic Benefits Per Trip vs. Base Case (2015)
Source: Johnston, Rodier, Choy and Abramham (2000).
This section highlights some findings on the individual and combined effects of transit, land use, and pricing scenarios using MEPLAN, an integrated transportation-land use model. Again, a much more detailed discussion and interpretation of results is provided in Johnston, Rodier, Choy, and Abraham (2000); the interpretations given below are drawn from this report.
Land use impacts - Identifying differences in the locations of households and employment between the baseline and evaluation scenarios is the primary reason for applying an integrated land use-transportation model. Some interesting differences can be noted based on the MEPLAN analysis:
Travel impacts - Under the HOV scenario, vehicle-trips are reduced slightly due to increased carpooling; however, VMT increases by about seven percent as a result of longer trip lengths made possible by increased highway capacity (Figure 12). The transit scenarios achieve maximum reductions in trips and VMT of roughly 10 percent; transit mode share increases significantly from 1.4 to 6.7 percent for the transit/TOD scenario. The addition of pricing has mixed effects; while it discourages development in TODs somewhat (as noted above), it also tends to reduce trip lengths.
Figure 8. HOV Scenario: Change in Employment
Johnston et al. (2000).
Figure 9 HOV Scenario: Change in Household
Johnston et al. (2000).
Figure 10. TOD, Light Rail, and Advanced Transit: Change in Employment
Johnston et al. (2000).
Figure 11. TOD, Light Rail, and Advanced Transit: Change in Households
Johnston et al. (2000).
Figure 12.
Transit and Land Use Scenarios: MEPLAN Model
Travel: Percent Change from Base Case (2015)
Source: Johnston, Rodier, Choy and Abraham (2000).
The study also compared the effects of high-occupancy roadway, land use, and pricing scenarios using the SACMET96 model. Except for the HOV-only scenario, these scenarios were not analyzed using the MEPLAN model. Nevertheless, the results are of interest because they illustrate the potentially synergistic effects of roadway, land use, and pricing scenarios, and they highlight the similarities and differences between HOV and HOT lane impacts.
Travel Impacts
The effects of the road-oriented policy scenarios in the Sacramento region are most strongly evident in the daily mode share projections for home-based work trips because these are the trips largely targeted by the HOT and HOV policies. It is in this trip purpose that congestion is most severe. For both the 2005 and 2015 scenarios, the HOV and HOT scenarios alone have a small impact on mode share. For the 2015 scenarios, the addition of the RODs to the HOT and HOV policies makes the mode share impact more noticeable. Drive-alone share decreases by about 1.5 percent, most of which is shifted to shared-ride. The addition of the pricing policies to the ROD scenarios further intensifies the effect, with a decrease in drive-alone mode share of up to four percent. The mode share trends for total regional trips are similar to those described above, but differences among scenarios are smaller because they include all trip purposes.
Daily vehicle travel projections for the 2015 scenario are presented in Figure 13. For both the 2005 and 2015 scenarios, the HOV and HOT scenarios result in small increases in vehicle trips VMT and greater reductions in VHD (5.2 to 7.7 percent). In the 2015 scenarios, the addition of RODs to the HOV and HOT scenarios produces an increase in vehicle trips, because of greater auto accessibility in the RODs, but reduces VMT. Thus, it appears that the ROD scenarios result in more but shorter vehicle trips. The addition of the RODs to the HOV and HOT scenarios significantly increases the reduction in VHD.
Figure 13.
Roadway and Land Use Scenarios: SACMET96 Model
Travel, Percent Change from Base Case (2015)
Source: Johnston, Rodier, Choy, and Abraham (2000).
The addition of the pricing policies to the HOV & ROD and HOT & ROD scenarios tends to dampen the increase in vehicle trips, increase the reduction in VMT, and increase the reduction of VHD, with a delay reduction of 23 percent in the HOT, ROD & Pricing scenario. In each case, the HOT option provides larger reductions in VHD than the HOV option. The HOT policies provide faster auto travel times, and thus lower VHD, than the HOV lanes because the tolls result in more efficient use of the new lanes.
Previously, it was found that the 2015 Light Rail scenario reduces vehicle trips by 0.1 percent, VMT by 0.3 percent, and VHD by 2.4 percent. Thus, with respect to reduction in VMT, the Light Rail scenario is superior to both the HOV and HOT scenarios, which increase VMT. However, with respect to reductions in delay the HOT scenario is best, followed by the HOV scenario, and then the Light Rail scenario. Considering land use and pricing policies, the same conclusions hold, although with significantly higher magnitudes of impacts in each case. The transit scenarios are preferable for reducing VMT, but the highway scenarios are preferable for reducing roadway delay.
Emissions Impacts
Impacts on daily vehicle emissions tend to correspond to VMT increases. Thus, both the HOV and HOT scenarios result in increased emissions over the base case scenario, and the HOT scenario produces slightly greater increases in emissions than the HOV scenario (Figure 14). Increased roadway capacity tends to speed up auto travel times, and thus auto travel distances (VMT) and emissions increase.
Figure 14.
Roadway and Land Use Scenarios: SACMET96 Model
Travel, Percent Change from Base Case (2015)
Source: Johnston, Rodier, Choy, and Abraham (2000).
The addition of the RODs to the HOV and HOT policies reverse the increase in emissions. However, the reduction in emissions is relatively small (approximately 2.0 percent for VMT and TOG, and slightly less for NOx). It was found earlier that TODs with similar regionwide shifts in new development and light rail and feeder bus service along similar corridors would produce a 6.5 percent reduction in VMT and a 5.6 reduction in TOG at the regional level. These results suggest that land use intensification policies may be more effective in reducing VMT and emissions when they are combined with transit than with HOV or HOT lane policies.
The addition of pricing policies to the HOV and HOT scenarios increases the reduction of emissions, although again, the reduction is smaller than for pricing in combination with transit and land use policies.
User Benefits
The change in economic benefits from the base case scenario (1995 present value) is shown in Figure 15. The HOT scenario produces larger economic benefits than does the HOV scenario ($0.01 versus $0.09 per trip for 2015). The HOV scenario results in a small economic loss when the full, unobserved cost of additional travel is included in the analysis. The HOT scenario results in an economic benefit because of the travel time saving to travelers with high values of time, which more than offsets the unobserved cost of additional travel.
Figure 15.
Roadway and Land Use Scenarios: SACMET96 Model
Economic Benefits Per Trip vs. Base Case (2015)
Source: Johnston, Rodier, Choy, and Abraham (2000).
In 2015, the addition of the RODs and then the pricing policies to the HOV and HOT scenarios both increase the economic benefits of the scenarios. This is because of the travel time saving that results from the RODs and the pricing policies. It is also assumed that revenues from the pricing policies are returned to the public (e.g., through lower sales taxes). Overall, the scenarios that combine RODs and pricing policies with HOT lanes have nearly double the benefits of the same policies combined with HOV lanes. The economic benefits for the HOT & ROD and the HOT, ROD & Pricing scenarios ($0.08 and $0.15, respectively) are comparable to those obtained from the Transit & TOD and the Transit, TOD & Pricing scenarios ($0.10 and $0.15).
Conclusions
HOV lanes and, to a lesser extent, HOT lanes are considered politically feasible policies in the Sacramento region to address the problems of congestion and emissions associated with regional transportation systems. The results of this study indicate that HOT lane policies may be somewhat better than HOV lane policies at reducing congestion. However, both the HOV and HOT scenarios increase VMT and emissions compared to a no-build scenario and the increase in emissions is greater for the HOT scenario compared to the HOV scenario. The Light Rail scenario is more effective at reducing VMT and emissions, but less effective at reducing congestion, than the HOV and HOT scenarios.
Combining land use intensification policies (RODs) and pricing policies with the HOT and HOV scenarios may produce added benefits with respect to congestion and emissions. The RODs and pricing policies are both found to reduce congestion, although more so in the HOT scenarios than in the HOV scenarios. The RODs and pricing policies are also found to decrease VMT and emissions; the benefits for these measures are small, however, especially when compared to TOD scenarios. These results suggest that land use intensification policies may be more effective at reducing VMT and emissions when they are combined with transit rather than with HOV and HOT lanes.
The results of this analysis indicate that HOV and HOT scenarios are not significantly different with respect to VMT, vehicle delay, and emissions; the HOT scenario, however, is clearly superior to the HOV and Light Rail scenarios as measured by user economic benefits. Economic losses for the HOV lanes are offset, and gains for the HOT lanes are increased, when the policies are combined with RODs and pricing policies. These increases in economic benefits are obtained from greater accessibility to carpooling and express transit service in the RODs and reduced congestion resulting from the pricing policies. The most beneficial roadway and land use scenarios provide similar economic benefits as the most beneficial transit and land use scenarios. In each case, these benefits are increased further when combined with pricing.
A comparison between the results produced by SACMET96 and MEPLAN suggests the contribution of land use modeling to the identification of travel impacts. As noted above, the scenarios simulated in SACMET96 differ somewhat from the scenarios simulated in MEPLAN. Thus, the contribution of the land use modeling cannot be fully isolated from the effects of the differences in the models and scenarios. However, some rough comparisons can be drawn:
The HOV-only scenarios are roughly comparable (Figure 16). While SACMET96 predicts an increase in total VMT, MEPLAN predicts a much larger increase. This may be a result of households and employment relocating to take advantage of the greater highway capacity and reduced travel times provided by the HOV-lane expansion.
The transit/TOD scenarios are roughly comparable (Figure 18), although (as noted above), the concentration of employment and population in TODs is not as great in MEPLAN as in SACMET96. MEPLAN produces a significantly greater mode shift from drive alone to transit. MEPLAN also forecasts a reduction in auto trips consistent with the mode shift, but the forecast reduction in VMT is smaller, suggesting that the remaining auto trips are longer on the average.
The magnitude of change in the policy scenarios compared to the base case is generally greater in the MEPLAN scenarios than in the SACMET96 scenarios. This may be due to the positive feedback between accessibility and land use in MEPLAN.
Figure 16.
SACMET96 vs. MEPLAN Model
Travel and Mode Share: Change from Base Case (2015)
Source: Johnston, Rodier, Choy, and Abraham (2000).
Figure 17.
SACMET96 vs. MEPLAN Model
Travel and Mode Share: Change from Base Case (2015)
Source: Johnston, Rodier, Choy, and Abraham (2000).
Figure 18.
SACMET96 vs. MEPLAN Model
Travel and Mode Share: Change from Base Case (2015)
Source: Johnston, Rodier, Choy, and Abraham (2000).
"The results of this study indicate that theoretically comprehensive urban models such as MEPLAN provide important insights into the development of heuristic policy strategies to address air quality problems." - (Johnston et al, 2000)
This case study illustrates both the advantages and limitations of applying a regional transportation-land use model. The level of effort involved in applying such a model is not inconsequential. At the same time, the results can provide important insights into the impacts of transportation policy on land use. The feedback of land use changes into travel, emissions, and other impacts may significantly affect these impacts compared to an assumption of static land use patterns.
The case study also illustrates the application of a typical travel demand model to an unusually broad range of policies. Some transit scenarios included the simulated effects of improved traveler information and demand-responsive suburban transit. "Roadway-oriented development" was analyzed in addition to transit-oriented development; HOV and HOT networks were combined with complementary policies to focus development around highway nodes. Pricing measures were also tested both individually and in combination with other actions. The pricing scenarios could represent policy decisions, but could also represent economic effects such as higher gasoline prices.
The results of the transportation and land use modeling described here, while interesting, are specific to the Sacramento area. Local results will depend upon a variety of factors such as development patterns, availability of land, and levels of roadway congestion and transit service. The study authors note, however, that some of their findings are consistent with results found in the literature. For example, other modeling and empirical studies in the U.S. have found that transit investment alone may provide relatively small reductions in VMT and emissions; impacts can be much more significant when combined with supportive land use policies and/or pricing measures.
The use of MEPLAN, or another transportation-land use model such as TRANUS or UrbanSim, provides a number of advantages for assessing regional transportation and land use policies. For example:
The inclusion of land use effects generally led to higher travel impacts (positive or negative) for comparable policy scenarios. The extent to which differences in models and scenarios contributed to this effect is unknown. It is possible, however, that much of this effect is due to the reallocation of population and employment in response to changes in travel times and costs. For example, in MEPLAN, HOV improvements led to increases in VMT that were disproportionate to trip and mode share impacts, suggesting a lengthening in average trip length.
Interest is growing in transit-oriented development, and a number of regions have modeled the travel impacts of concentrating development in transit station areas. The actual mechanisms for influencing development, however, are generally not identified. A land use model such as MEPLAN, which simulates land markets, can be used to test different policy mechanisms for achieving development objectives. In this application of MEPLAN, it appeared that it would be difficult in practice to achieve the desired level of reallocation. For example, no combination of "reasonable" tax incentives and subsidies in MEPLAN was sufficient to obtain densities in transit zones as high as those manually allocated to transit zones in the SACMET96 model.
The results are not always intuitive and are by no means linear combinations of individual policies. For example, in this study, parking pricing tended to discourage development from TOD zones, thus offsetting tax incentives.
Models such as MEPLAN, TRANUS, and UrbanSim have a strong economic foundation, and business and residential location decisions are modeled based on a range of factors. Land markets are key to the modeling framework, with prices driving development. This is in contrast to models such as DRAM-EMPAL that reallocate land use only on the basis of transportation accessibility. UrbanSim advances the state-of-the-practice by using the logit model structure, typically used to predict transportation mode choice, to model a broad range of decisions such as business and residential location.
"Large-scale urban models are best used as heuristic policies guides, that is, for suggesting direction and magnitude of change and rank ordering of scenarios." - (Johnston et al, 2000)
Like most transportation-land use models, MEPLAN uses large zones and sketch networks, potentially leading to different travel and emission results than the Sacramento model which is much more detailed. Specifically, the representation and accuracy of projected vehicle volumes, speeds, distances, and thus emissions, is limited. In the Sacramento case, SACOG staff have expressed concern over using two different models that produce different results. If land use feedback is to be incorporated into travel modeling for regional planning, conformity analysis, or other purposes, an approach will be required that uses a land use model in iteration with a travel model that includes a detailed network and zone system.
The tradeoff for a detailed representation of land markets and inter-industry interactions is that a large amount of data is required to develop and calibrate the model. While MEPLAN can be customized based on locally available data, many regions will not have adequate data to take full advantage of the modeling capabilities offered. The application in Sacramento would have been strengthened through the availability of floorspace data as well as data from multiple years. The project demonstrates, however, that a useful model of urban land use and transport interaction can be developed in a United States metropolitan area using existing data sources.
Behavior is modeled at an aggregate level (groups of people or businesses by zone) rather than at a disaggregate or individual level. This masks differences in behavior among individuals or subsets of the population. While the aggregate approach is less realistic, it is much simpler than modeling an entire heterogeneous market.
It is possible to develop an integrated travel demand model with finer geographic detail into the MEPLAN framework; however, its development may be time consuming due to the difficulties of calibrating such a model. Recent advances in calibration methods may address this problem and reduce the time and monetary cost of model development (Abraham and Hunt, 2000). In addition, land use models can be developed for use with typical regional travel demand models. In many regions in the U.S., however, travel demand models would need to be significantly improved to better represent travel time and cost throughout the model hierarchy.
There is currently no "goods movement" flow in the Sacramento version of MEPLAN, due to a lack of local data. The model has been designed to include this at a later date, however. Inclusion of freight flows would improve the accuracy of the model's firm-location choice and would allow transportation conditions for freight to be investigated separately from conditions for passenger travel.
In future research, the study authors plan to control for the travel model in MEPLAN and thus better isolate the contribution of the land use component of the model.
UrbanSim, developed in the U.S. and applied in the Honolulu, Portland, and Salt Lake City metropolitan areas, provides a number of advances in the state-of-the-practice in urban transportation-land use modeling. These include, for example, disaggregation of the location choice to the level of traffic analysis zones; integration with existing four-step travel models; and the option for parcel-level microsimulation of land development and redevelopment. UrbanSim allows explicit input of public sector choices as policy scenarios, and further expands the set of land use policies that can be tested.
The user benefit analysis illustrates how an important effect - overall benefits to travelers - can be measured and compared among scenarios. Simply measuring time savings for existing trips does not provide a complete picture of traveler benefits. The measure used here also considers changes in travel cost, other measures that affect the overall "utility" of travel, benefits to new travelers, and the overall capital and operating costs of infrastructure. In addition, the measure calculates benefits differently for commercial vehicle travel than for personal travel. As a result it provides a picture of whether the project is economically beneficial overall (although external costs and benefits are not included).
User benefits can further be analyzed by income group, consistent with income groupings used in the travel model. This can help in analyzing the equity impacts of various policy combinations. For example, the pricing/transit scenario and the TOD/transit scenario were both found to have positive net economic benefits. The pricing/transit scenario, however, affected the lowest-income group negatively, while the TOD/transit scenario affected all income groups positively.
Abraham, J. E., and J. D. Hunt (1998). "Calibrating the MEPLAN model of Sacramento." Presented at the 77th Annual Meeting of the Transportation Research Board, Washington D.C., January 1998, Paper No. 980649.
Abraham, J. E., and J. D. Hunt (1999). "Firm location in the MEPLAN model of Sacramento." Transportation Research Record 1685.
DKS Associates (1994). Model Development and User Reference Report. Prepared for Sacramento Area Council of Governments, Sacramento, CA.
Johnston, Robert A.; Caroline J. Rodier, Melanie Choy, and John Abraham (2000). Air Quality Impacts of Regional Land Use Policies: Final Report for the Environmental Protection Agency. Department of Environmental Science and Policy, University of California - Davis, February 2000.
Rodier, Caroline J.; John E. Abraham and Robert A. Johnston (2000). "Air Quality Analysis of Transportation: Is it Important to Model the Land Use Effects?" Presented at the 79th Annual Meeting of the Transportation Research Board, Washington, D.C., January 2000, Paper No. 00-1118.
Organization |
Person |
Phone |
|---|---|---|
University of California - Davis |
530-582-0700 |
|
530-757-2791 |
||
University of Calgary (MEPLAN application) |
403-230-5897 |
|
Douglas Hunt |
||
U.S. EPA |
202-260-5447 |
