Overall, the AMBAG travel demand model meets current state-of-the practice travel modeling standards and has some special features that go beyond standard practice. The model contains detailed socioeconomic data and has a geographically based TAZ, link and node structure. Trips are modeled by purpose and the model uses common trip types. The model uses standard 4-step modeling practices including a gravity-based distribution model, a nested-logit mode choice model, a transit assignment model, and an equilibrium-based highway assignment model. The AMBAG travel demand model also contains a congestion feedback loop between distribution, mode choice and highway assignment to account for congested travel times in trip distribution.
The AMBAG travel demand model contains many additional features to be more relevant to answering questions specific to the region:
The AMBAG travel model uses a 2005 base year and was estimated, calibrated and validated using local survey and count data.
The AMBAG model is designed primarily for use in transportation planning at a regional scale. As such, the AMBAG regional travel model has been designed primarily for the following purposes:
Because the model is designed and calibrated for macro-level regional planning activities, uses of the model at a more detailed level may require localized model validation and updates to serve project-level applications. The AMBAG regional travel model has also been used for the following detailed planning efforts:
The following sections identify issues and concerns highlighted by AMBAG and their partner agencies during the TMIP peer review. In addition, AMBAG provided a list of specific questions for the peer review panel prior to the peer review meeting, which is included in Appendix D of this report.
4.2.1 California Law SB 375 Implementation
Senate Bill 375 was passed into California law requiring the need for a process for reducing greenhouse gas (GHG) emissions through more sustainable land use and transportation planning. The legislation aligns regional land use, transportation, housing, and GHG reduction planning efforts. AMBAG as an MPO is required to develop and adopt the Metropolitan Transportation Plan (MTP) and Sustainable Communities Strategies (SCS) pursuant to SB 375 beginning with the adoption of AMBAG's 2013 plan with final compliance required by the adoption of their 2017 plan.
Implementation of SB 375 will require regional coordination in addition to modeling various complex policies and having extensive public participation. Public participation also entails further developing visualization and interactive modeling tools.
California commissioned a statewide household travel survey (CHTS) in 2001-2002. This survey was supplemented by two add-on surveys totaling 1,564 surveys in the AMBAG region representing 0.69% of the households (total households from 2000 Census). Another CHTS is expected to be conducted in 2011-2012 with an expected sample size in the AMBAG region of 1,316 households. This represents 0.56% of the households (households from 2005 ACS).
AMBAG is concerned that the sample size of the data being collected for the three county AMBAG region is too small for full use in the travel demand model. They are considering adding onto the 2011-2012 survey to increase the sample size to achieve a sample rate of at least 1% of households in the MPO region. AMBAG would like to know if this is a good use of funds and if just increasing the sample size is sufficient or if oversampling special markets or under represented demographic groups would be better.
4.2.3 Agency Resource Constraints
AMBAG expressed concerns about the balance between the uses and demands of the travel demand model and the amount of resources allocated to running, maintaining and developing the model. AMBAG has the equivalent of 0.75 full time staff who can work on the travel model on a regular basis. In addition, few funding streams have been identified for model development and data acquisition and processing. AMBAG would like advice from the peer review panel on ways to achieve better balance in this arena and still maintain a realistic schedule of model improvements that meet the needs of model users and to implement SB 375 requirements.
4.2.4 Member Agency Requests of the AMBAG Model
The three RTPAs and Caltrans are the most frequent users of the AMBAG travel demand model. TAMC would like to use the AMBAG model for a light rail study that would be seeking FTA New Starts funding and for a bus rapid transit study. They question whether the current version of the AMBAG model would be able to pass FTA's rigorous standards. The RTPAs also expressed an interest to study alternative modes (including bicycle), shifting agriculture goods movement from truck to rail and to test land use policy using the '5D' tool. Caltrans expressed concerns that the current AMBAG travel model is not calibrated to peak period conditions, which is a key aspect of some of their planning work. They would also like to see thought given to development of a meso-scopic model and a dynamic traffic assignment.
In addition, the two transit districts in the AMBAG region would like to be able to use the travel model to better understand transit markets and where to invest transit services. Information they wish to see out of the model include an origin and destination of transit riders, where to provide service, choice vs. captive riders, and assessing demand responsive service and select patronage service (college campuses, military bases or tourists).
To be more useful to their planning purposes, these entities expressed a desire for a better mode choice component of the model. There was also a general concern of how peak period congestion was (or was not) being addressed in the travel model. There was a desire expressed for better representation of peak periods due to long commuting patterns as some locations are experiencing extended AM & PM peak periods or nontraditional peaks from students and service workers.
AMBAG has developed a draft model improvement plan (MIP) to address concerns with the model and to provide continuing support to maintain a travel model that is representative of current state-of-the-practice. This plan emphasizes the development of integrated land use and transportation models with various customized modeling and visualization tools to analyze the impact of specific development (residential, commercial, industrial and mixed-use developments) and their impacts on VMT and resultant GHG emissions. The proposed AMBAG MIP is also designed to enhance the modeling capabilities to accomplish SB 375 requirements as well as provide support in developing and planning for Sustainable Communities Strategies (SCS) throughout the Monterey Bay region. The following are the main priorities in the proposed AMBAG MIP:
Short Term Improvements for 2013 MTP
Medium Term Improvements for 2017 MTP
Long Term Improvements for future MTPs
The following sections summarize the information provided by AMBAG staff as well as comments from peer review participants. AMBAG provided the peer review panel with model documentation describing all of the model components, input data and model validation results.
4.4.1 Household Travel Survey and Transit On-Board Survey Data
The AMBAG regional travel demand model used data from the 2001-2002 California Household Travel Survey (CHTS) to calibrate trip generation cross-classification rates and mode choice constants and to validate the distribution model's friction factors and trip length frequencies. However as discussed in the concerns section of this report, the MPO expressed the need to add to the sample size for their area beyond what is being considered in the 2011-2012 CHTS. The MPO stated that an add-on to the survey was to be a set amount if no specialty target groups were identified, but the cost would be more if these groups were to be identified.
The transit agencies servicing the AMBAG region collect on-board surveys on a regular basis. However, these surveys are being collected for the transit agencies' purposes and not for travel model estimation and development. There is no uniform transit on-board survey across the transit districts.
Panel Discussion
The panel recommended that the MPO add to the 2011-2012 CHTS sample size for future model improvements. So much of the model's components and parameters depend on household survey data that making this data set more robust would be valuable. The MPO would also see an economy of scale by piggy-backing onto the CHTS. The panel commented that oversampling specialty groups such as low-income-migratory or high-income-retiree populations would be valuable.
Oversampling will not necessarily achieve the data for the transit model parameters. The panel suggested AMBAG use a transit on-board survey for model validation and recommended working with transit providers to conduct a comprehensive on-board survey.
4.4.2 Highway and Transit Networks
The AMBAG regional travel model uses a road network based on county street centerline GIS data. The 2005 base year and 2035 forecast horizon networks were both prepared based on the 2000 base year network. The road network for Santa Cruz and Monterey counties includes all public roadways. For San Benito and Santa Clara counties, only the more important streets (major collector and above) are represented. The 2035 network was built from the 2005 network and includes the projects in the AMBAG MTP. However, improvements to the local system identified in general plans are not included in the 2035 network. The highway network has a single centroid connector to each TAZ in most cases.
Free flow speeds and capacities are coded onto the network links. Free flow speeds were developed from empirical data collected summer of 2002 and 2003 and included a broad sampling of all freeways and arterials and a select sampling of collector and local roads. Free flow speeds are cross classified by functional type and area type. Capacities are estimated based on Highway Capacity Manual 2000 procedures and link-based geometric and operational characteristics. Link capacities are calculated as hourly service flows then factored by 2.7 for the 3-hour peak period capacity and 12.0 to represent daily capacity. The model then uses volume to delay functions to calculate congested speed based on volume/capacity ratios. The volume to delay functions were based on research conducted by MTC and represent modified versions of the standard BPR curves.
The transit network consists of a description of bus lines that are superimposed on the road network. Transit line characteristics include the locations of stops, walk access links, the peak and midday headways, and the average travel speeds of buses to the average speed of all vehicles on each type of road. Transit lines are coded in the model for the following services:
Panel Discussion
The panel expressed concern that the model is using average daily speeds from the travel time speed runs. Average daily speeds will cloud congestion. Also the panel questioned if the average daily speeds are representing the free flow speeds in the volume delay functions. If congested speeds are being used then the speeds representing the start of the functions are too low and would affect travel time calculations in the model's path skimming.
The panel also discussed the high level of network detail, specifically the inclusion of all local roads in Santa Cruz and Monterey counties. Most panel members agreed that the AMBAG model is small enough where the extra detail would not cause an extraordinary burden and in fact, the added detail was useful for modeling accuracy. Some panel members felt that this level of detail was unnecessary and only served to increase model run time and time to code the network. It was suggested that the detail could be left in as an input and toggled off during the model run if performance became an issue. The panel also wondered if the local roadway detail was acting as pseudo centroid connectors for these two counties as the documentation indicated there tended to be only one centroid connector per TAZ and recommended adding more centroid connectors.
4.4.3 Traffic Analysis Zone Structure and Socioeconomic Data
The traffic analysis zone (TAZ) structure in the AMBAG model is based on Census 2000 block geography. Zones range in size from 0.03 sq. mi. to 394 sq. mi., but the typical TAZ in an urban area is approximately 0.15 sq. mi. TAZ boundaries were drawn with a desire to add as much detail as possible such that non-motorized and walk access to transit trips could be modeled properly. TAZs for the AMBAG model cover the three AMBAG counties of Santa Cruz, Monterey and San Benito as well as all of Santa Clara County. Santa Clara County TAZs were derived from the Santa Clara Valley Transportation Authority (VTA) model and the AMBAG TAZ structure nests within the VTA TAZ structure. The Santa Clara TAZs are at a more aggregate scale except for the southern part of the county along the Hwy 101 corridor. The AMBAG model has 1,884 internal TAZ and 10 external TAZ.
Land use and socioeconomic data are coded at the TAZ level. The AMBAG model uses households stratified by four age categories, four income categories and four auto ownership categories. These disaggregate values were taken directly from the 2000 Census. A household stratification model is not used. Socioeconomic data also includes average household size. In forecast years, each TAZ's income and auto ownership stratification assumptions are held constant whereas in future scenarios the number of households in each age group changes. Employment data is represented by six employment categories based on SIC classifications. Employment allocation used a detailed employer sample from Info USA and expanded by employment category based Department of Finance county control totals (earlier versions of the model used Woods and Poole Economics to obtain employment expansion factors). Farm employment was allocated into five farm product types using parcel data and a consultation with regional Farm Bureau representatives. Socioeconomic data for Santa Clara County was taken from the VTA travel model.
The TAZ data also contain the following information to support two school trip purposes and a visitor trip purpose in the AMBAG model:
Panel Discussion
The panel liked that there were variables in the input data, such as hotel rooms, to capture visitor trips. The panel asked how enrollment data was forecast. The MPO responded that K-12 enrollment used what was reported by each school district. If the district projected new schools or school closures, this was reflected in the data. University enrollment used 20-year estimates from the universities. The panel also suggested looking at providing detail on special markets, such as military personnel and full and part-time residents (migrant workers and retirees are main part-time populations but are at different ends of income spectrum). Details to consider could include location as well as season when the part-time population is present.
AMBAG develops population, housing and employment forecasts for Monterey, Santa Cruz and San Benito counties. Population forecasts used a cohort-component model and included details such as age, gender, birth/death rates and in/out migration. Housing unit forecasts pivot on projected population, group quarters, average household size and vacancy rates. Employment forecasts use an input-output economic model. These methods determine county level control totals that are then compared with California State county-level estimates. After negotiations, a singular set of county-level control totals is agreed upon. County level totals are then disaggregated to local jurisdictions based on trend forecasting and planning assessment of accessibility to services, amenities and jobs as well as input from a committee of local jurisdictions. The final socioeconomic forecast is then adopted by the AMBAG board.
Panel Discussion
AMBAG would like to update their current socioeconomic forecasting process for a parcel-based, integrated land use / transportation model and asked the panel to comment. The panel offered that there is a wide spectrum of land use models, ranging from more simple models like UPLAN to more complex models like UrbanSim and PECAS. It was suggested that several mid-range land use models, such as DELTA and Cube Land, require considerably less investment to build and maintain (approximately a fifth the cost) than the more complex land use models. These mid-range models can be designed to be TAZ-based but have a parcel data option and are less data hungry than the more complex models. The panel recommended investing in the more mid-range complexity land use model, especially in light of the resource constraints of the MPO. The panel also recommended that a new land use model could be built to recognize seasonal residents and workers and that it did not necessarily need to be fully integrated or automated with the travel model to include a transportation element.
The panel also commented on the experience of several areas that are implementing land use models. Puget Sound Regional Council has successfully implemented an UrbanSim model and is happy with it at the regional level. They are just starting to use it at the county or local level. The Denver Regional Council of Governments has also tested UrbanSim and concluded that it worked well in some areas and not so well in others, particularly the development model side which tended to be more idiosyncratic. The panel suggested that this issue also existed in PECAS model applications. Two of the MPOs in the San Joaquin Valley are implementing Cube Land as part of their model improvement program.
The panel inquired to understand better the reasons behind the need for a new land use model. Among other needs, the MPO stated that SB 375 required the use of a parcel-based GIS process to be used in forecasting. The panel stated that the MPO did not need a fully integrated land use / transportation model to fill the GIS-parcel requirement. However, AMBAG would like to have an end game in mind so when data sets are developed, such as in the blueprint planning process, they are developed using the correct framework. They would also like an automated process of linking parcel data with the travel model, which is currently a manual process.
AMBAG stated that they have much parcel level data collected and commented that some county data are not yet in electronic format. Assessor data are available but there is a cost and time lag associated with these data. The panel recommended that AMBAG define a process for developing parcel data through a regional data consortium. Local communities could feel a benefit from sharing data and working together by receiving information for smart growth concepts such as the '5Ds' analyses to minimize carbon footprint and VMT. This data sharing could enrich the quality of the databases and provide a cost sharing benefit.
Trip generation rates are organized by seven basic trip purposes:
Home-based Work trip production rates are stratified by the four household age classifications and four income groups, resulting in 16 rates. Home-Maintenance, Home-Discretionary and Other trip production rates are stratified by four age classifications and four auto ownership groups. The Other trip purpose also includes an additional generation based on employment by type of employment. Home-based School trip production rates are based on four age classifications. Trip attraction rates are calculated based on various combinations of employment categories, K-12 and university enrollment data, age classification and auto ownership classification. Visitor trips are based on attraction locations and are considered the only special generators in the AMBAG model. Trip generation rates were derived from the 2001-2002 CHTS, which included Monterey, San Benito and Santa Cruz counties and the 2002 TAMC/SBCOG survey, which was conducted using the same survey methodology as the 2001-2002 CHTS.
The current AMBAG model has a freight purpose and is viewed as a placeholder which adds significant value over the alternative of accounting for commercial vehicle travel under the Other trip purposes. The truck model was developed using Quick Response Freight Manual (QRFM) version 1 trip rates to generate regional truck trips. An initial distribution was estimated using a traditional urban model type gravity approach (QRFM friction factors) which was then modified with an automated matrix estimating routine (ODME) to create a base year Truck OD matrix. The process produced truck assignments that are consistent with base year truck counts. Future truck amounts were estimated using the same method then were used to factor the base year ODME matrix to produce a future truck matrix.
Panel Discussion
The panel made several comments regarding the trip generation model. The panel noticed that the trip rates for K-12 seemed low. They recommended reporting unadjusted and adjusted production to attraction balances, which would clarify this and other issues related to the production and attraction equations. The panel also recommended smoothing trip rates. The panel also asked why income was used in trip generation. The MPO did not know as this was inherited from previous work. The panel commented that the there seemed to be a casual relation between age and trip rates and suggested that household size would show a more direct relationship and recommended using household size, autos or income and workers as variables and estimating trip rates for entire households, not per person in the household. The age of head of householder variable will be more useful at the household level than at the person level.
The panel also questioned whether the trip generation model at the TAZ level cross classifications would be sensitive enough to capture the '5Ds' in post processing or it needed to be done at the parcel level.
The panel also asked about the military presence in the region. They wondered how people living on the military base were being handled in the model and the nature of their trip making along with other activities associated with the military institutions. It was suggested the MPO consider making the military base a special generator in the AMBAG model.
The panel discussed the current method of using ODME to develop freight trip tables and suggested that this method was not helpful for future planning, since it is insensitive to future investments or economic indicators. The panel suggested that other methods be considered for local truck travel and the new statewide freight model be considered for external truck travel.
4.4.6 External Travel and Linkage to the California Statewide Model
The model is using a methodology where external-internal (EI-IE) trips are incorporated into the internal trip distribution process. EI-IE trips are expressed in terms of productions and attractions for each trip purpose by model gateway using the following process:
Panel Discussion
The peer review panel discussed the use of the California statewide model in determining external trip flows. The California statewide model is currently being developed. A base year is finished and future years are under development. The statewide model could provide interregional and through trips for regular trip purposes as well as recreational trips and goods movements. The panel asked if the AMBAG model and the statewide model had consistent TAZ and network data. The MPO responded that they provided these inputs to Caltrans and that the statewide model was used in the derivation of external trips in the AMBAG model. However, the actual use of the statewide model was unclear in the documentation. Because the statewide model is focused on long distance travel, the panel recommended that AMBAG use the statewide model for external trips and to begin discussion with other planning entities.
The trip distribution model uses a gravity model to estimate trip interactions between zones. Average daily congested travel times are used for impedance values and friction factors are set using a gamma function. The gamma function parameters for each trip purpose were calibrated using the trip length frequencies obtained from the estimated congested zone-to-zone travel times of the survey trips and the calibration tools included with the TransCAD software. The productions and attractions are doubly constrained for all but the School, University and Visitor trip purposes that are constrained to attractions. The current version of the AMBAG model does not use K-factors.
The distribution model used self-reported trip times by trip type from the linked trip household survey data set for validation. Work trip distribution was compared to the detailed CTPP data for year 2000. Because of the differing analysis years, the commute patterns were expressed in terms of percent of originating trips and percent of workplace trips.
Panel Discussion
The panel stated that trip distribution will be hard to capture in the AMBAG region with a gravity model. For example, the validation histograms for Home-based Work trips show modeled trips as smooth curves that taper off as travel times increase but the observed data has distinct bumps at longer travel times representing commuting trips between communities. The model results should also be bumpy to reflect the commuting patterns of isolated cities. The gravity model alone will not fix this. The panel discussed ways to address this issue. The panel suggested that in the short term AMBAG reinvestigate the strategic use of a limited number of K-factors.
The panel also recommended looking into a destination choice model to address the distribution issue. This could be in conjunction with a scaled implementation of an activity based model.
In addition, the panel suggested investigating how the model is handling short vs. long distance trips. The panel asked what happened to the distribution results when Santa Clara County was added. They recommend reporting screen line validation at the Santa Clara County boarders to check. The panel also recommended sub-dividing trip length frequency reporting by county to better assess the validation of the trip distribution model.
The panel commented on the use of stated travel times from the household survey in validation. Typically travel time as reported in a household survey is inaccurate due to an incorrect perception on how long a trip actually takes or because people tend to round their times. A better approach is to assign the observed origin and destination from the survey to the model network to get a better sense of actual travel times. The panel recommended fixing the observed travel times before validating the trip distribution model. This should also be completed after the recommendations about the highway assignment and networks are completed.
The AMBAG model uses a nested logit choice model for determining the mode of each person trip. The choices are grouped so that choices in the same level have similar sensitivities to travel characteristics. The logit model has the following nests:
Level 1 Motorized vs. Non-Motorized
Level 2 Motorized: Automobile vs. Transit
Level 3 Automobile: Drive Alone vs. Shared Ride
Level 3 Transit: Walk Access vs. Drive Access
Level 4 Shared Ride: 2-Person vs. 3+ Persons
Level 4 Walk Access: Premium Transit Service vs. Local Transit Service
Level 4 Drive Access: Park and Ride vs. Kiss and Ride
The most current version of the model was updated to reflect FTA guidance for New and Small Starts forecasting. The previous calibration to the 2001-2002 CHTS was done prior to the issuance of the guidance and thus contained coefficients that were out of accepted ranges and had county specific variables that are no longer allowed. Only coefficients outside the FTA parameters were adjusted (to be the mid-point of the range of acceptable values) and modal constants were also modified to more closely match the observed targets. Previously developed mode targets by percent share originally derived from the 2001-2002 CHTS were utilized for this process.
Panel Discussion
The panel asked whether the MPO was considering congestion pricing as this is not a choice in the current mode choice model. The MPO responded that congestion pricing was not critical in their current planning but that HOV was important because it is being considered in Santa Cruz County.
The panel also commented on several facets of the mode choice documentation. The documentation did not report at which level nesting coefficients are applied. The documentation also reported just the mode choice constants. Good practice is to report these constants in terms of equivalent minutes of in-vehicle time. The panel also commented that the source for the driving and parking costs used in the transit skimming process were not included in the documentation. They recommended that all these items be added to the documentation. The model documentation also suggested that improvements have been made to bring the logit model's parameters into line with FTA's standards. However, certain panel members disagreed and felt that the mode choice bias constants were still too high for use in any new starts analysis.
The panel noted that the mode choice parameters were inconsistent with the transit skim weights. Inconsistencies between the two can cause problematic results. The panel recommended providing consistency between the mode choice coefficients and transit skims.
The panel recommended recalibrating the mode choice model in light of the short term trip generation, distribution and mode choice fixes. They also recommended a long term improvement of estimating new mode choice models with socioeconomic and demographics variables that are sensitive to urban design and level of service variables for use in the '5D' analyses.
4.4.9 Highway and Transit Assignment
The AMBAG regional travel model uses a multiclass, origin based user equilibrium assignment. The origin based user equilibrium assignment method was used because it allows a very high degree of convergence. The four vehicle classes being assigned are:
Transit trips are assigned to the AMBAG transit network separately based on each of the four types of transit service and access:
The transit assignment assumes that all persons within each of these four categories will use the shortest available time path. No capacity constraints or multiple assignment iterations are used for the transit trips. The total ridership on each transit line is calculated by adding the results of the four individual transit assignments.
The AMBAG model is currently set up to run highway and transit assignments for five separate time periods:
Trips by period were tabulated using the 2001-2002 CHTS.
Highway assignment was validated against 2005 vehicle count data from 527 locations. This data was assembled from the various AMBAG jurisdictions and Caltrans District 5. The on-going traffic monitoring programs of the three RTPA's are the primary count database sources. All counts were 24 hour counts. Peak period data were not available and there has been no validation against hourly or peak period count data. Transit boarding data was obtained from two of the three transit agencies providing fixed route bus service in the three county AMBAG region. No detailed boarding data was available for San Benito County. Transit data for Santa Clara County were not used in the validation process. Transit assignment was validated using 2005 boarding data.
The 2005 base year highway assignment validated well when compared to 527 count locations around the region. The model validation against counts shows an overall 29.1% RMSE and -1.9% count VMT error. The system-wide modeled 2005 base year VMT estimate is consistent with the 2005 HPMS estimate (within -5%). The model also closely replicated total ridership by each transit system, but individual route results vary widely. The overall transit validation had a 40.75% RMSE and an aggregate loading ratio of 0.98. The overall AMBAG model validation results fall within FHWA/FTA acceptable ranges (RMSE 40% or less and HPMS VMT +/- 5%).
Panel Discussion
The panel discussed the convergence of the assignment network and noted that the documentation does not indicate what convergence threshold is used. The panel recommended checking and reporting the assignment convergence and suggested the tolerance should be 10E-6.
The panel confirmed that the model is being validated to actual counts for volume and to HPMS for VMT. Caltrans noted that they are in the process of installing more ATR's in the AMBAG region as part of their traffic monitoring program. AMBAG also commented on the AMBAG use of many screenlines to validate the model. The panel suggested that perhaps the MPO does not need to use as many screenlines in their validation process.
The panel commented on the type of assignment method used and the use in regards to select link analyses. Though the origin based user assignment converges more quickly than other methods, select link analyses require an assignment method that maintains proportionality. The 'check proportionality' option should be used in TransCAD if origin-based assignment is used, or another assignment process that guarantees proportionality (such as Frank-Wolfe) should be used for select link analyses.
The panel also cautioned against running 24-hour assignment and suggested assigning each period separately then adding to a daily total. They also recommended using five time periods (AM, midday, PM, evening and night) to better account for certain trips that are of interest to the MPO, such as hospitality service trips. These periods could be recalculated from the CHTS survey data. The panel also noted that the documentation needs to report how trip purposes were combined for estimating time of day factors.
In the discussion, concern was raised by a member agency that 2005 boarding data used to calibrate the base year transit contained anomalies that affected ridership, including a strike. The panel recommended that instead of using a single validation year in which the data may be experiencing volatility, a multi-year trend based validation would be more appropriate and defensible. In the case of transit ridership, boardings from 2004 and 2006 could also be used to determine an appropriate validation target. The peer review panel also noted that since the household survey was conducted in 2001-2002, system performance data from that time period should be used to match the observed survey data. The MPO stated that 2002 was used to estimate mode share data but the transit assignment was developed later and used 2005 data. The model base year is 2005 and all data sets (socioeconomic, traffic counts, transportation networks, etc.) represent 2005 conditions.
4.4.10 Congested Travel Time Feedback Loop
The AMBAG regional travel model uses a distribution, mode choice, assignment feedback loop. In the loop congested travel times from assignment are brought back as inputs to trip distribution and mode choice. The loop is executed two times. For the first pass, congested travel speeds are estimated based on factors applied to the coded free flow speeds on each link. These estimated congested speeds are used as input to the first pass trip distribution and mode choice steps. The second pass calculates congested speeds on each link based on the traffic assignment completed during the first pass. The feedback loop is intended to ensure that the final congested travel speeds used for air quality analysis are consistent with the travel speeds used throughout the model system.
Panel Discussion
The panel recommended that the MPO consider using more than two iterations to obtain congested travel times. Reporting the convergence results can help determine the preferred number of feedback iterations.
AMBAG used the regional travel model for air quality determination. AMBAG is currently in attainment for all air quality pollutants.
Panel Discussion
The panel asked if AMBAG expected to become non-attainment for any air quality pollutants in the future. The MPO responded that a small pocket of San Benito County will become non-attainment for PM2.5 when the standard is released. The greater concern for the MPO is providing modeling to support greenhouse gas emission requirements. The panel commented that modeling for air quality conformity can drive various aspects of the model and model validation.
The panel had a lengthy discussion on the staff and financial resources available to the MPO to run and maintain the travel modeling program. As identified in the "Concerns Identified by AMBAG and Member Agencies" section of this report, AMBAG has the equivalent of 0.75 full time staff committed to running the travel model program.
The panel expressed concern with the level of staffing and the MPO's ability to accomplish the model improvements needed to support the 2013 MTP requirements. The panel stressed that uses of the model should drive how the model should be built and the complexity of the data. As many demands are being placed on the AMBAG model, the panel recommended that AMBAG push for additional resources to accomplish modeling goals. The panel recommended the MPO leverage partnerships in the region, such as with other planning agencies or with the universities. Data sharing could help reduce cost and provide some of the needed resources. In addition, the panel recommended leveraging existing grants for improving modeling process and exploring other grant sources. The panel also recognized the need to increase funding for model improvements. Because there is a need by many for good travel forecasts, the broad customer base should recognize they have a stake in the development of the model as well as a need to pitch in with funding or staff.
The panel felt that with some improvements the AMBAG travel model would be able to support the 2013 MTP and begin to meet the requirements of SB 375, though it may not incorporate all of the SB 375 requirements in this timeframe. During the model development process, the AMBAG regional model can continue to be used for planning applications with the understanding that model deficiencies are being worked on and the model user needs to assess the reasonableness of the model for a particular application.
4.4.13 Release of the AMBAG Regional Travel Demand Model
AMBAG expressed concern about releasing the full model and having the model results misrepresented. AMBAG felt this could place them in the situation of having to defend outside uses of the regional travel demand model.
The panel recommended AMBAG be transparent in its operations. Having an open process puts the MPO on the 'good guy list' and gives the MPO less liability because users can't claim they didn't know. The panel suggested that the responsibility to defend should be on the model user or applier and not the MPO. The panel recommended AMBAG use a user agreement similar to that in other urban areas. The model user would be responsible for changes made to the model, its application and results. Changes and official version releases can be handled or tracked with versioning software.
4.4.14 Activity-based Travel Models
There was a lot of discussion on activity based models. The panel discussed the value of using an activity-based model in this area but recognized that full development of an activity-based model would be difficult due to cost and resource constraints. The panel talked about the value of transferring an activity-based model, such as from Lake Tahoe, Fresno, San Joaquin, Seattle or the California statewide model. Some panel members thought that transferring from a smaller activity based model could be accomplished in about a year for about a third the cost of a full independent development. Other panel members expressed concern with transferring models from other areas due to the fact that a thorough process is often not followed and cited research from Chandra Bhat. All agreed that the second generation activity-based models will come with a reduced cost as the first generation models had a steeper learning curve and have established a knowledge and user base from which to build. If AMBAG switched to an activity-based model, they would be in the second generation. It was also suggested that AMBAG could do a phased approach to activity-based model implantation, similar to the Puget Sound Regional Council. A hybrid activity trip based model could use daily activity patterns and primary destination choice and auto ownership models and could leverage the statewide model population synthesizer. The panel also expressed concern that the MPO might not have resources to run or maintain an activity based model.
