RIVCOM Mode Choice Models
Introduction
This section covers the mode choice model and post mode choice components prior to highway assignment. Post mode choice components include directionality and person trip to vehicle trip conversion step.
Mode Choice Model
The mode choice model converts person trips created by Trip Generation and in Production-Attraction trip tables by Trip Distribution to person trips by mode. This model is only applicable to the model area residents. Trucks, IE/EI and EE trips are already in vehicle trip units, so there is no need for a mode choice for those segments.
Mode Choice Model Specification
Mode choice models are mathematical expressions that are used to estimate the modal shares of the travel market given the time and cost characteristics of the various competing modes, the demographic and socioeconomic characteristics of the residents, and the unincluded attributes of the modes represented in the model. Mode choice models are typically modeled using one the three types of Logit models: multinomial logit, hierarchical logit and nested logit. The RIVCOM mode choice model used a nested logit structure. The standard logit formulation can be expressed as:
The utility expression for each available mode i is specified as a linear function which incorporates a range of variable types, including time, cost, locational measures, and the socio-economic characteristics of the traveler. For example:
RIVCOM Mode Choice Model
The RIVCOM mode choice model is a nested logit model. The mode choice structure is shown here:
Travel modes are grouped in separate nests based upon similarities in their attributes. This structure consists of several nests with a total of 9 mode choice alternatives. The top level nest classifies person trips into auto, transportation network company (TNC, ie. Uber, Lyft, etc.), transit and non-motorized trips.
- The auto nest classifies trips into drive alone mode and shared ride. Shared ride includes the shared ride 2 (2 occupants in vehicle) and shared ride 3 plus (3 or more occupants in vehicle) modes in the third nest.
- The TNC nest has only one mode: ride hail mode.
- The transit nest classifies trips into three access modes: walk, park-and-ride (PNR) and kiss-and-ride (KNR).
- The non-motorized nest classifies trips into walk and bike modes.
The RIVCOM model evaluates mode alternatives for a total of 13 trip purposes including five HBW purposes where work trips are further stratified by income and auto sufficiency, and non-work trip purposes as listed below:
- Home-based work
- Zero autos
- Low income insufficient (income <= $40k where autos = 1 & workers > 1)
- Low income sufficient (income <= $40k where autos = 1 & workers = 1, or autos greater than 1)
- High income insufficient (income > $40k where autos = 1 & workers > 1)
- High income sufficient (income > $40k where autos = 1 & workers = 1, or autos greater than 1)
- Home-based other
- Zero autos
- Low income insufficient (income <= $40k where autos = 1 & hh size > 1)
- Low income sufficient (income <= $40k where autos = 1 & hh size = 1, or autos greater than 1)
- High income insufficient (income > $40k where autos = 1 & workers > 1)
- High income sufficient (income > $40k where autos = 1 & hh size = 1, or autos greater than 1)
- Non-home-based
- School
- University
The following abreviations are used in the model outputs to tag the segments fpr HBW and HBO and will also be used later in the model summaries.
| Segment Name | Segment Abreviation |
|---|---|
| Zero autos | v0 |
| Low income insufficient | ilvi |
| Low income sufficient | ilvs |
| High income insufficient | ihvi |
| High income sufficient | ihvs |
Mode Choice Parameters
The RIVCOM mode choice model parameters were asserted from other models such as SCAG Activity-based Model, and represent generally accepted industry standard coefficients for regions without sufficient data for a mode choice model specification. The in-vehicle time (IVT) coefficient was asserted to be 0.025 and the cost coefficient was pivoted from the IVT coefficient based on the median household income of different household segments, the details are presented in the section “Wage Rates and Cost Coefficients” later. The mode choice model parameters by purpose and market segments are available here: mode_choice/mc_parameters
Logical and consistent nesting coefficients are applied at each mode choice nesting level. The RIVCOM model procedures neither use transit networks nor generate transit skims; however the mode choice model allows the user to calibrate transit trips by access mode to observed data. A transit availability matrix was used to define origin destination zone pairs with possible transit trips. During calibration these availability indicators were set to zero for zone pairs which have no transit boardings, in SCAG 2020 RTP base year (2016) model output, in the corresponding Community Statistical Area (CSA) geography pair. CSA stands for Community Statistical Area number around 300. They are one of the geographic classification maintained in the SCAG model. Users can modify transit availabilities for three access modes including walk, PNR and KNR to transit; however, currently the availability matrix is set to be same for all the access modes. These transit availability matrices are located here in the file _trn_availability.mtx. mode_choice/availability_matrix
Wage Rates and Cost Coefficients
Purpose specific cost coefficients are derived from the value of time values in the RIVCOM region. The average household income values were computed from the ACS PUMS 2014-2018 five year data. Subsequently, the value of time for each trip purpose and income group combination was computed from the Riverside and San Bernardino county mean income values. The region mean household income value is $85,765 for the two county region in 2018 dollars. The value of times were calculated as per standard modeling practices and used in the RIVCOM model are:
- Home-based work
- Zero autos ($7.76/hr)
- Low income insufficient ($5.18/hr)
- Low income sufficient ($4.23/hr)
- High income insufficient ($19.73/hr)
- High income sufficient ($22.15/hr)
- Home-based other
- Zero autos ($3.88/hr)
- Low income insufficient ($2.59/hr)
- Low income sufficient ($2.12/hr)
- High income insufficient ($9.87/hr)
- High income sufficient ($11.08)
- Non-home-based ($8.25/hr)
- School ($3.88/hr)
- University ($7.76/hr)
Cost coefficients (VOT divided by the IVT coefficient) by purpose and market segments are available in the mode choice parameter file. The ‘auto_op_cost’ term is the cost coefficient and can be found here: mode_choice/mc_parameters
School Bus Trips
School bus trips are removed from the trip distribution output trip tables prior to mode choice model application. The share of school bus trips was calculated as 11% from the SCAG ABM model outputs for the school purpose, and this value is used in the RIVCOM model as well. The current value of the school bus mode share parameter is here. mode_choice/school_bus_share
Mode Choice Calibration
The calibration targets for mode choice were prepared using SCAG 2020 RTP base year (2016) model output. Only trips that had at least one trip end, origin or destination, were included in creating the calibration targets. Targets were created for each of the 13 purpose market segment combinations. During calibration, the model was also summarized using only trips with at least one end within Riverside County. The tables below show the comparison between calibration targets and model outputs.
| Trip Purpose | Trip Segment | Drive Alone | Share Ride 2 | Share Ride 3 | Walk Transit | PNR Transit | KNR Transit | Walk | Bike | Ride Hail |
|---|---|---|---|---|---|---|---|---|---|---|
| HBW | v0 | 0 | 40 | 34 | 3 | 0 | 0 | 8 | 3 | 2 |
| HBW | ilvi | 60 | 16 | 16 | 0 | 1 | 0 | 4 | 2 | 1 |
| HBW | ilvs | 69 | 16 | 9 | 0 | 1 | 0 | 3 | 1 | 1 |
| HBW | ihvi | 78 | 11 | 6 | 0 | 0 | 0 | 3 | 0 | 0 |
| HBW | ihvs | 83 | 9 | 4 | 0 | 1 | 0 | 2 | 0 | 0 |
| Trip Purpose | Trip Segment | Drive Alone | Share Ride 2 | Share Ride 3 | Walk Transit | PNR Transit | KNR Transit | Walk | Bike | Ride Hail |
|---|---|---|---|---|---|---|---|---|---|---|
| HBW | v0 | 0 | 43 | 30 | 11 | 0 | 1 | 8 | 4 | 3 |
| HBW | ilvi | 59 | 13 | 18 | 1 | 1 | 0 | 5 | 2 | 1 |
| HBW | ilvs | 69 | 12 | 10 | 2 | 3 | 0 | 2 | 1 | 1 |
| HBW | ihvi | 79 | 9 | 7 | 1 | 0 | 0 | 2 | 1 | 0 |
| HBW | ihvs | 83 | 9 | 5 | 1 | 1 | 0 | 1 | 0 | 0 |
| Trip Purpose | Trip Segment | Drive Alone | Share Ride 2 | Share Ride 3 | Walk Transit | PNR Transit | KNR Transit | Walk | Bike | Ride Hail |
|---|---|---|---|---|---|---|---|---|---|---|
| HBO | v0 | 0 | 16 | 33 | 11 | 0 | 0 | 30 | 4 | 7 |
| HBO | ilvi | 33 | 28 | 24 | 1 | 0 | 0 | 11 | 2 | 1 |
| HBO | ilvs | 35 | 29 | 21 | 1 | 0 | 0 | 11 | 2 | 1 |
| HBO | ihvi | 41 | 28 | 19 | 0 | 0 | 0 | 9 | 1 | 1 |
| HBO | ihvs | 41 | 27 | 20 | 1 | 0 | 0 | 9 | 1 | 1 |
| Trip Purpose | Trip Segment | Drive Alone | Share Ride 2 | Share Ride 3 | Walk Transit | PNR Transit | KNR Transit | Walk | Bike | Ride Hail |
|---|---|---|---|---|---|---|---|---|---|---|
| HBO | v0 | 0 | 26 | 33 | 8 | 0 | 0 | 23 | 4 | 6 |
| HBO | ilvi | 32 | 30 | 24 | 1 | 0 | 0 | 10 | 2 | 1 |
| HBO | ilvs | 35 | 31 | 21 | 1 | 0 | 0 | 10 | 2 | 1 |
| HBO | ihvi | 44 | 29 | 18 | 0 | 0 | 0 | 7 | 1 | 1 |
| HBO | ihvs | 44 | 28 | 18 | 0 | 0 | 0 | 8 | 1 | 1 |
| Trip Purpose | Trip Segment | Drive Alone | Share Ride 2 | Share Ride 3 | Walk Transit | PNR Transit | KNR Transit | Walk | Bike | Ride Hail |
|---|---|---|---|---|---|---|---|---|---|---|
| HBSC | all | 18 | 30 | 28 | 1 | 0 | 0 | 20 | 2 | 1 |
| NHB | all | 42 | 30 | 23 | 0 | 0 | 0 | 3 | 0 | 1 |
| HBU | all | 63 | 17 | 8 | 1 | 1 | 0 | 5 | 3 | 1 |
| total | All | 45 | 26 | 19 | 1 | 0 | 0 | 8 | 1 | 1 |
| Trip Purpose | Trip Segment | Drive Alone | Share Ride 2 | Share Ride 3 | Walk Transit | PNR Transit | KNR Transit | Walk | Bike | Ride Hail |
|---|---|---|---|---|---|---|---|---|---|---|
| HBSC | all | 8 | 36 | 28 | 1 | 0 | 0 | 20 | 5 | 1 |
| NHB | all | 48 | 28 | 19 | 0 | 0 | 0 | 3 | 0 | 1 |
| HBU | all | 59 | 17 | 10 | 4 | 3 | 0 | 4 | 3 | 1 |
| All | All | 44 | 26 | 18 | 1 | 0 | 0 | 8 | 2 | 1 |
Trip Directionality
For all the steps in the RIVCOM model flow up to and including mode choice model, the trips are in “Production/Attraction” (PA) format. This simply means that all home-based trips start at home and end somewhere else. This simplification of reality is done for several reasons in the model but is not actually how travel occurs. As a result, before highway trip assignment can take place, the PA format must be converted to Origin/Destination" (OD) format. In this format, a trip from work to home (considered a home-based work trip) starts at work and ends at home.
This conversion is accomplished using factors stratified by time of day and purpose. As an example, the majority of HBW trips in the AM start at home and end at work. As a result, the PA-to-OD factor (“PA factor” for short) is typically greater than 50% (> .5). In the PM period, this trend is usually reversed, and the PA factor is typically less than 50% (< .5).
The PA factors used in RIVCOM were asserted from the 2012 SCAG regional model parameters for home based purposes for the three time periods. By definition, non-home-based trips are the same in either format as there is no “home” end of the trip; their PA factors are set to .5. The same treatment is applied to commercial vehicles, trucks, and external truck trips. For external passenger trips (IE/EI), there was slight asymmetry. Only 47% of trips in AM period was in the direction of PA (Model area to outside model area). Similarly for other time periods also there was slight asymmetry for external passenger trips (IE/EI).
The directionality factors are applied after the mode choice models. The RIVCOM model directionality factors are here: directionality/directionality_factors. The table below shows the current directionality factors used in the model.
| Purpose | AM | OP | PM |
|---|---|---|---|
| HBW | 0.970 | 0.602 | 0.063 |
| HBO | 0.901 | 0.578 | 0.380 |
| HBSC | 1.000 | 0.096 | 0.032 |
| HBU | 0.976 | 0.523 | 0.369 |
| NHB | 0.500 | 0.500 | 0.500 |
| NHBNR | 0.500 | 0.500 | 0.500 |
| CV | 0.500 | 0.500 | 0.500 |
| SUT | 0.500 | 0.500 | 0.500 |
| MUT | 0.500 | 0.500 | 0.500 |
| IEEI | 0.470 | 0.506 | 0.510 |
| IEEI_Trk | 0.500 | 0.500 | 0.500 |
Vehicle Occupancy Factors
In addition to applying directionality factors before assignment, the person trip matrices are converted to vehicle trips using occupancy factors. These factors were calculated from the 2012 SCAG regional model person and vehicle trips, available in the validation report. The vehicle occupancy factors used in RIVCOM model are here: directionality/veh_occ_factors. A summary of those factors are also shown in the table below. Note that the specification file veh_occ_factors.csv, allows for the specification of occupancy factors differently for each market segment for a purpose; however, the occupancy factors are set to the same value for different segments in the RIVCOM model (e.g. low income vehicle insufficient, low-income vehicle sufficient etc.).
| Purpose | Mode | AM | OP | PM |
|---|---|---|---|---|
| HBW | da | 1.00 | 1.00 | 1.00 |
| HBW | sr2 | 2.00 | 2.00 | 2.00 |
| HBW | sr3 | 3.51 | 3.51 | 3.51 |
| HBW | ride_hail | 1.00 | 1.00 | 1.00 |
| HBO | da | 1.00 | 1.00 | 1.00 |
| HBO | sr2 | 2.00 | 2.00 | 2.00 |
| HBO | sr3 | 3.55 | 3.55 | 3.55 |
| HBO | ride_hail | 1.20 | 1.20 | 1.20 |
| HBSC | da | 1.00 | 1.00 | 1.00 |
| HBSC | sr2 | 2.00 | 2.00 | 2.00 |
| HBSC | sr3 | 4.62 | 4.62 | 4.62 |
| HBSC | ride_hail | 1.20 | 1.20 | 1.20 |
| HBU | da | 1.00 | 1.00 | 1.00 |
| HBU | sr2 | 2.00 | 2.00 | 2.00 |
| HBU | sr3 | 3.55 | 3.55 | 3.55 |
| HBU | ride_hail | 1.20 | 1.20 | 1.20 |
| NHB | da | 1.00 | 1.00 | 1.00 |
| NHB | sr2 | 2.00 | 2.00 | 2.00 |
| NHB | sr3 | 3.64 | 3.64 | 3.64 |
| NHB | ride_hail | 1.20 | 1.20 | 1.20 |
Riverside County Model, 2020