3 Building blocks
3.1 Household synthesis and land use
One of the strengths of VisionEval is the rich detail on individual households. This allows for household specific policies, travel behavior can respond to specific household costs and attributes, and outputs can be mined for differences by population groups. The approach of building on a synthesized population borrows from the state of the practice in activity-based travel demand models.
Watch a short video on household synthesis
VisionEval takes user input statewide population by age group, assembles them into households with demographic attributes (lifecycle, per capita income) and allocates them to BZone-level dwelling units. Separately, BZones are attributed with employment and land use attributes (location type, built form ‘D’ values, urban mixed use, and employment by type). Household members are identified as workers and/or drivers and the number of household vehicles are estimated. Each home and work location is tied to a specific Bzone with its associated attributes.
Policies are added to each household as a function of their home and work Bzones:
- Parking restrictions (work and non-work)
- Travel demand management (TDM) programs (home and work-based)
- CarService program availability
The following sections describe each module that contributes to this concept.
3.1.1 Synthesize households
Household synthesis is carried out in several steps:
- Create customized PUMS dataset: This is done prior to setting up a model in VisionEval. A household dataset is prepared from Census PUMS data for the modeled region. The default data included with VisionEval is for Oregon. PUMS data for other regions may be used instead, rebuilding the package to reflect Census households for the region of interest.
- Create Households: The identified types of PUMS households are expanded to meet user control totals and other demographic inputs. Census PUMS data are used define probabilities that a person by age group would be found in each of hundreds of household types. A matrix balancing process is used to allocate persons by age to each of the PUMS household types in a way that matches input control totals and optional constraints. The sampled households are expanded to meet user control totals and other demographic inputs.
- Predict Workers: The number of workers by age group within each simulated household is predicted using Census PUMS probabilities.
- Assign LifeCyle: Categorizes households are categorized by six lifecycle categories given the household age mix and employment status.
- Predict Income: The annual income for each simulated household is predicted as a function of the household’s worker count by age group, the average per capita income where the household resides (AZone), and interactions between neighborhood income and age (all and seniors). The models are estimated with Census PUMS data.
3.1.2 Household drivers and autos
The number of drivers and autos in each household are calculated in two steps:
- Assign Drivers: Drivers by age group are assigned to each household as a function of the numbers of persons and workers by age group, the household income, land use characteristics, and transit availability. Metropolitan areas are also sensitive to transit service level and urban mixed use indicators at the home location. Optional restriction on drivers by age group can be used in calibration or to address trends such as lower millennial licensure rates.
- Assign Vehicle Ownership: The number of vehicles owned or leased by each household are determined as a function of household characteristics, land use characteristics, and transportation system characteristics. Households in metropolitan areas are also sensitive to transit service level and urban mixed use indicators at the home location. The model first predicts zero-auto households and then the number of vehicles owned (up to 6), if any.
3.1.3 Land use attributes
Two steps are required to add land use attributes to the synthetic population:
- Calculate 4D Measures: Several land use 5D built form measures are calculated for each Bzone. The
destinationaccessibilities are based on Bzone population, employment, dwelling units, and
developable land areainputs. The
designvariable is a user input.
- Calculate Urban Mixed Use Measure: An urban mixed measure for the household is calculated based on population density of the home Bzone and dwelling unit type. The model is based on 2001 NHTS data. The model iterates to match an optional input target on the share of households to locate in urban mixed-use areas.
3.1.5 Policy levers
Several assumptions about parking, demand management, and mobility services can also be coded:
Assign Parking Restrictions: Households are assigned specific parking restrictions and fees for their residence, workplace(s), and other places they are likely to visit based on parking inputs by BZone (within Bzones coded as within metropolitan areas [Marea] only).
- Residential Parking Restrictions & Fees: The number of free parking spaces available at the household’s residence is set based on input value that identify the average residential parking spaces by dwelling type in each Bzone. For household vehicles that cannot be parked in a free space a residential parking cost (part of auto ownership costs) is identified as a function of input parking rates for the home Bzone (if any).
- Employer Parking and Fees: Which workers pay for parking is set by inputs that define the proportion of workers facing parking fees in each Bzone. Whether their payment is part of a cash out/buy back program is similarly set by input proportions by Bzone and associated fees set by input parking rates for the work Bzone.
- Non-work Parking Fees: The cost of parking for other activities such as shopping is estimated as the likelihood that a household would visit each Bzone and the parking fee in that Bzone. The likelihood is calculated with a gravity-type model, given the relative amount of activity in the Bzone (numbers of households by Bzone and the scaled retail and service job attractions by Bzone as marginals) and the proximity to each destination (inverse distance matrix from home Bzone seed matrix). The average daily parking cost is a weighted average of the fee faced in each destination bzone and the likelihood of visiting that Bzone.
Assign Demand Management: Households are assigned to individualized marketing programs based on input participation levels within their home Bzone. Each worker in the household can also be assigned to an employee commute options program based on input participation levels for workers within their assigned work Bzone. A simple percentage reduction in household VMT is applied based on the household’s participation in one or more of these program (maximum of multiple program participation, to avoid double-counting). Worker reductions are only applied to that worker’s work travel portion of overall household VMT, and summed if multiple workers in the household participate in such programs.
Caution: The model assumes high-caliber TDM programs are in place that produce significant VMT savings. Inputs should reflect this.
Assign CarSvc Availability: A car service level is assigned to each household based on the input car service coverage for where the household resides (Bzone). High Car Service availability can have an impact on A auto ownership costs (which affects number of autos owned by the household) and auto operating cost (see the discussion in the next chapter on household costs and budgets).
3.2 Household multimodal travel
Watch a video overview of the Household Multimodal Travel module
Travel of various modes by households (vehicle, transit, bike, and walk modes) is estimated as a simple function of the rich demographic and land use attributes of the household. In metropolitan areas travel is also influenced by inputs on transport supply on a per capita basis, such as available roadway capacity and bus-equivalent transit service levels. Traditional travel models incorporate behavioral dynamics in choice models to build tours and trips for each synthetic person. VisionEval, by contrast, uses simple regression equations that directly estimate average per capita trips and miles by mode, linked by average trip lengths.
After adjusting VMT for household budget limitations it is further adjusted for household participation in TDM programs (home & work-based) and short-trip SOV diversion before calculating household trips for all modes. The household’s bike miles are also adjusted to reflect SOV diversion input.
The following sections describe each module, which are implemented in sequence:
- The household’s daily VMT is calculated with household budget adjustments
- The vehicle operating costs are calculated
- The vehicle operating costs are adjusted to fit within the
- Daily VMT reductions due to TDM measures and short-trip SOV diversions are calculated
- Vehicle and non-vehicular (AltMode) trips are calculated for each household
3.2.1 Transport supply
Transport supply variables are defined in two steps. Note that these calculations are only carried out within metropolitan areas (Mareas) only:
- Assign Transit Service: Transit service levels are input for each metropolitan areas and neighborhood (Bzone). Annual revenue-miles (i.e. transit miles in revenue service) by eight transit modes are read from inputs for each metropolitan area. A Bzone-level Transit D attribute defines access to transit (not yet work access) for each household based on inputs on relative transit accessibility. Using factors derived from the National Transit Database (NTD), input annual transit revenue miles for each of the eight transit modes are converted to bus-equivalent miles by three transit vehicle types (van, bus, and rail). Per capita relative transit supply and bus-equivalent service-miles are calculated.
- Assign Road Miles: Stores input on the numbers of freeway lane-miles and arterial lane-miles by metropolitan area and year. Computes the relative roadway supply, arterial and freeway lane-miles per capita.
3.2.2 Household travel calculations
Household travel by vehicles are calculated in three steps:
- Calculate Household Daily VMT (Dvmt): Household average daily vehicle miles traveled (VMT) is estimated as a function of household characteristics(income, workers, children, drivers), vehicle ownership, and attributes of the neighborhood (population density) and metropolitan area (urban mixed-use, transit service level, road lane-miles) where the household resides. It also calculates household VMT percentiles which are used by other modules to calculate whether a household is likely to own an electric vehicle (EV) and to calculate the proportions of plug-in hybrid electric vehicles (PHEV) VMT powered by electricity. First, households with no VMT on the travel day are identified. Then VMT is estimated for those that travel. Average and VMT quantiles are estimated to reflect day-to-day variance that helps identify whether an EV vehicle is feasible for this households typical travel. These values are derived from the 2001 NHTS data.
- CalculateVehicleTrips: This module calculates average daily vehicle trips for households consistent with the household VMT. Average length of household vehicle trips is estimated as a function of household characteristics (drivers/non-driers, income), vehicle ownership (auto sufficiency), and attributes of the neighborhood (population density) and metropolitan area (urban mixed-use, freeway lane-miles) where the household resides, and interactions among these variables. The average trip length is divided into the average household VMT to get an estimate of average number of daily vehicle trips.
- Calculate AltMode Trips: This module calculates household transit trips, walk trips, and bike trips. The models are sensitive to household VMT so they are run after all household VMT adjustments (e.g., to account for cost on household VMT) are made. Twelve models estimate trips for the three modes in metropolitan and non-metropolitan areas, in two steps each. The first step determines whether a household has any AltMode trips and the second part determines the number of trips. All of the models include terms for household characteristics (size, income, age mix) and the household’s overall VMT. Neighborhood factors (population density) factors into all but the bike trip models. For households in metropolitan areas transit service level has an impact as well, with transit ridership also sensitive to when residents live in urban mixed-use neighborhoods.
3.2.3 SOV diversion
Household single-occupant vehicle (SOV) travel is reduced to achieve bike and micro-transportation input policy goals, i.e., for diverting a portion of SOV travel within a 20-mile tour distance (round trip distance). This allows evaluating the potential for light-weight vehicles (e.g. bicycles, electric bikes, electric scooters) and infrastructure to support their use, in reducing SOV travel. First, he amount of the household’s VMT that occurs in SOV tours having round trip distances of 20 miles or less is estimated. Then the average trip length within those tours is estimated. Both models are sensitive to household characteristics(drivers, income, kids), vehicle ownership (auto sufficiency), and attributes of the neighborhood (population density, dwelling type) and metropolitan area (urban mixed-use, freeway lane-miles) where the household resides, and the household’s overall VMT. Both models have multiple stages, including stochastic simulations to capture day-to-day variations.
The diversion of these short trips is assumed to only apply in urban and town location types. The VMT reductions are allocated to households as a function of the household’s SOV VMT and (the inverse of) SOV trip length. In other words, it is assumed that households having more qualifying SOV travel and households having shorter SOV trips will be more likely to divert SOV travel to bicycle-like modes. The estimates of the household’s share of diverted VMT, average trip length of diverted VMT are applied elsewhere to reduce DMVT and increase bike trips. Zero vehicle households are not allowed to divert SOV travel. Census PUMS data is used to estimate the models.
3.2.4 DVMT reductions
Each household’s VMT is adjusted for their TDM program(s) participation, if any, as well as input from metropolitan area short-trips SOV diversion goals. The SOV diversion also increases bike trips (diverted SOV VMT divided by SOV average trip length).
3.3 Vehicles, fuels and emissions
The powertrains, fuels, and associated emissions datasets for all modeled vehicle groups are among the most complex inputs used in VisionEval. Default datasets are included in the VisionEval installer to simplify this for the user. The user can use these defaults or develop their own data that matches the VisionEval input requirements. It is anticipated that different datasets will be developed by users that can be shared with the VisionEval community. For example, one package may represent a base scenario of federal vehicle, fuel, and emission standards, while another package represents the California zero-emissions vehicle (ZEV) rules and low carbon fuel for the home location’s CarService.
The model looks in household vehicle sales tables indexed by vehicle type and age to determine the probability of each powertrain in that sales year, along with its associated fuel efficiency and other attributes. Each household vehicle is assigned attributes consistent with these probabilities. In some cases electric vehicles (EVs) are replaced by plug-in hybrid electric vehicles (PHEVs) if household VMT and residential charging limitations exist.
The powertrain mix of non-household vehicle groups – car service, commercial service, transit, and heavy trucks – is applied to VMT (rather than individual vehicles) in the scenario year (rather than sales year). There is some input adjustment for average vehicle age and commercial vehicle type share.
Fuels for each vehicle groups can rely on the package defaults, or use one of two input options. The user can either provide a composite carbon intensity representing all gallons of fuel used for that vehicle group, or provide fuel mix shares (base fuel mix, biofuel blend proportions), combined with package default lifecycle (well-to-wheels) carbon intensity for the individual fuels. The resulting carbon intensity per gallon are applied to gallons generated from VMT and vehicle fuel efficiency assumptions. Adjustments to fuel efficiency due to reduced speeds due to congestion as well as ITS operational programs (e.g., speed smoothing) and EcoDrive programs.
The table below summarizes the vehicle and fuel options available within VisionEval.
|Vehicle Group||Vehicle Types||Powertrain Options||Veh Input Adjustments||Fuel Options|
|Household Vehicles||automobile, light truck||ICE, HEV, EV, PHEV||(default veh mix), age, %LtTrk||gas/ethanol, diesel/biodiesel, CNG/RNG|
|Car Service VMT||automobile, light truck||ICE, HEV, EV||veh mix, age (HH %LtTrk)||gas/ethanol, diesel/biodiesel, CNG/RNG|
|Commercial Service VMT||automobile, light truck||ICE, HEV, EV||veh mix, age, %LtTrk||gas/ethanol, diesel/biodiesel, CNG/RNG|
|Heavy Truck VMT||heavy truck||ICE, HEV, EV||veh mix||gas/ethanol, diesel/biodiesel, CNG/LNG|
|Public Transit VMT||van, bus, rail||ICE, HEV, EV||veh mix||gas/ethanol, diesel/biodiesel, CNG/RNG|
Note that individual vehicles are modeled for households, based on sales year default datasets and age of the owned vehicle. Other groups’ vehicle and fuel attributes apply to VMT in the scenario modeled year. As a result, PHEVs do not exist other than household vehicles, instead PHEVs are represented as miles driven in HEVs and miles in EVs.
Watch a video overview of vehicles, fuels, and emissions
3.3.1 Household vehicle table
The household vehicle table is generated in two steps:
- Create Vehicle Table: A vehicle table is created with a record for every vehicle owned by the household, and additional vehicle records are added to reach the household’s number of driving age persons. Each vehicle record is populated with household ID and geography fields (Azone, Marea) and access time attributes. Each vehicle record is either “own” or (worker without a vehicle) assigned access to a CarService level, depending upon coverage in the household’s home Azone.
- Assign Vehicle Type:. Identifies how many household vehicles are light trucks and how many are automobiles as a function of number of vehicles, person-to-vehicle and vehicle-to-driver ratios, number of children, dwelling unit type, income, density, and urban mixed use data (in metropolitan areas only).
3.3.2 Powertrains and fuels defaults
These values are defined in two steps:
- Load Default Values:. This script, run before the rest of VisionEval is started, reads and processes the default powertrains and fuels files in the package and creates datasets used by other modules to compute fuel and electricity consumption, and well as associated fuel and electricity carbon intensity emissions from vehicle travel.
- An Initialize step is run by VisionEval as part of its initialization on each run. Optional user-supplied vehicle and fuel input files, if any, are processed (including input data checks). When available, modules that compute carbon intensities of vehicle travel will use the user-supplied data instead of the package default datasets.
3.3.3 Assign household powertrains and fuels
The powertrain and fuel type is assigned to each vehicle in each household in three steps:
- Assign Vehicle Age: Assigns vehicle ages to each household vehicle and CarService vehicle used by the household as a function of the vehicle group (household vehicles only), household income, and assumed mean vehicle age by vehicle type and Azone. The age model starts with an observed vehicle age distribution and relationship between vehicle age and income. These data are currently based on summaries of the 2001 NHTS. Adjustments are made based on user average vehicle age inputs (household by vehicle type, car service overall).
- Assign Household Vehicle Powertrain: This module assigns a powertrain type to each household vehicle. The age of each vehicle is used with default tables by vehicle type that identify the powertrain mix of vehicles sold in each sales year. Other default tables identify vehicle characteristics tied to powertrain that include battery range, fuel efficiency, and emissions rate. Assignments of EVs may be changed to PHEVs if the battery range is not compatible with estimated day-to-day trip lengths, or the home dwelling lacks vehicle charging availability.
- Calculate Carbon Intensity: This module calculates the average carbon intensity of fuels (grams CO2e per megajoule) by vehicle group and, if applicable, vehicle type. Average fuel carbon intensities for transit vehicle modes are calculated by metropolitan area, other vehicles are calculated for the entire model region. The module also reads the input average carbon intensity of electricity at the Azone level.
3.3.4 Assign non-household powertrains and fuels
The assignment of powertrain and fuel characteristics is carried out in two steps:
- Calculate Transit Energy And Emissions: This module calculates the energy consumption and carbon emissions from transit vehicles in urbanized areas. Assumptions (package default or user input) on powertrain mix and fuels for three transit vehicle types by metropolitan area are applied to associated Marea transit service miles for these types. Assumptions (package default or user input) on average carbon intensity of fuel and electricity by transit vehicle types are then applied to Marea fuel and electricity usage across types to calculate carbon emissions.
- Calculate Commercial Energy And Emissions: The energy consumption and carbon emissions of heavy trucks and commercial service VMT (no vehicles) are calculated by on-road (not sales) year. VMT shares of Commercial Service powertrains by vehicle type and heavy truck powertrains are calculated (per package default or user input). Any fuel efficiency (MPG and MPKWH) adjustments are then applied, due to policies (EcoDriving, speed smoothing and/or congestion (including effects of any ITS operational and congestion fee policies). Ecodriving applies only to internal combustion engine (ICE) vehicles and ITS operational policies and congestion apply only in metropolitan areas. Both vary by powertrain and for commercial vehicles, vehicle type. Combining fuel efficiency and VMT (from the Household Multimodal Travel Model) results in estimates of energy usage (fuel and electricity). Fuel carbon intensity for these modes is calculated by metropolitan area and/or region and applied to fuel and electricity usage to estimate CO2e emissions.