The purpose of this scope of work is to develop the next generation of the BKR travel demand model in cooperation with the Cities of Bellevue, Kirkland, and Redmond. The next generation BKR model will be a localized implementation of PSRC’s SoundCast model, which uses DaySim for demand modeling and EMME for supply modeling.

TASK 1: PROJECT MANAGEMENT AND ADMINISTRATION

Progress meetings via conference call shall be conducted on a bi-weekly basis. The Consultant shall provide the City with an agenda before each progress meeting. The Consultant shall be responsible for taking summary notes at all meetings and shall supply the City with a copy of the notes.

A Project schedule shall be developed by the Consultant to establish Baseline Start and Baseline Finish dates for all tasks and deliverables under this AGREEMENT, and shall be submitted to the City for review and approval one week after the first model design workshop. The project schedule shall show a critical path leading to the project completion date. The completion date shall be no later than June 30, 2017. The schedule for this project shall be updated monthly and submitted with each Progress Report. The updated schedule shall show Baseline, Actual and/or Projected Start and Finish dates that reflect the actual progress of the project. The Consultant shall prepare and submit a Progress Report with each invoice.

Deliverable(s): Conference call management, meeting notes, progress reports, schedule updates; Due throughout the project

TASK 2: FINALIZE MODEL DESIGN

Consultant will meet onsite with the BKR model development team to finalize the model design. The design should strive to keep the BKR and SoundCast code implementations synched up as much as possible. It is assumed that PSRC will participate in the meeting as well. Ahead of the meeting, Consultant will prepare a draft recommended model design memo for review. Key model design requirements to finalize include:

• Fully document the extent to which random seed variation affects the results of all sub-models within DaySim. To meet BKR modeling needs for stable arterial volume, develop and program an approach to systematically control and minimize random variation. This includes, but not limited to how best to implement zonal-based over- and under-sampling of households in order to reduce runtimes while increasing the stability of results. Consultant will evaluate over-sample BKR households by a factor of between 3 to 5, while at the same time, under-sample households outside BKR at a rate that decreases w.r.t. their proximity to BKR. The sample rate will be specified at the TAZ level and the sampling will be done by household income and size category in order to ensure a representative sample. The expanded trips output by DaySim will then be aggregated into assignment matrices, thereby “averaging” the results. The Consultant will also investigate the merit and impact of removing random variations outside the BKR area for potential implementation.
• Simplifications to the transferred SoundCast model in order to reduce runtimes and data storage space, without significantly compromising results. Options to be evaluated and tested include, but are not limited to:
  • Reduce 12 time period skims to fewer time periods
  • Simplify the toll/non-toll by value-of-time group to just toll eligible by value-of-time group.
  • Identify features in model run stream that can be turned on and off to suit different model applications such as development impact review, land use scenario planning, adding new roadway facilities, and roadway closures.
  • Identify and evaluate ways to speed up the calculation of iterations through increased processing power or coding efficiencies.
• Model improvements:
  • Depending on the performance of the transferred and revised SoundCast model, test and implement measures to improve adopted BKR model performance in time-of-day and mode choice models.
  • Incorporate PSRC’s ongoing model development work within the project schedule and budget. PSRC is planning to 1) calibrate and re-estimate the model using the 2014 household survey and 2015 university travel survey, 2) add bike assignment, and 3) update the supplemental demand models trucks, EI - IE, and special generators. Depending on the state of these, as well as PSRC’s schedule for more completion, Consultant will discuss with the model development team the timing and how best to integrate these improvements.
  • Add taxi/ride-sharing as its own mode in the mode choice model as recommended by the Peer Review Panel. This will give the analysts flexibility later on to do scenario testing, such as different level of market penetration of autonomous vehicles. Data to develop the model can be asserted for the time being since there may not be enough data to properly estimate the coefficients.

Deliverable(s): One-day Onsite Workshop, Model Design Memo, Due 4 weeks from NTP

TASK 3: IMPLEMENT MODEL IMPROVEMENTS

Consultant will first test the revisions specified in the model design memo into the existing draft BKR version of SoundCast, which is the PSRC version but with the old BKR zone system and networks. Consultant will implement revisions into the new BKR zone system and networks when it is available. The BKR SoundCast model is run through the same Python script as PSRC’s version - run_soundcast.py in the command console. Network skimming, running of DaySim, running of additional demand models, and network assignment is fully automated. PSRC’s synthetic population procedures will be used as well for population synthesis .Consultant will develop and implement the model improvements as follows:

• Develop, upon approval of the City, implement a more user friendly frontend routine for preparing land use data inputs
• Evaluate and document how random seed variation affects the results of all the choice models within DaySim. Recommend, and upon the approval of the City, implement the sampling and stability improvements and integrate these into the overall model run stream
• Identify and implement the simplifications to the setup of the revised BKR zone system and networks
• In coordination with PSRC’s SoundCast on-going model improvement efforts, adopt or design and implement the followings:
  • Bike assignment routine either within EMME or GIS environment
  • More consistent integration of supplemental demands including trucks, EI - IE, through, and special generators trips for assignments.
  • Diagnose and remove potential coding errors in SoundCast before or during adoption.

Consultant will create a new private project site on GitHub and check-in the initial version of the model setup. The model system will be revised in iterations and available to the BKR model development team via GitHub. Consultant will document the model system with the linked GitHub project wiki using GitHub Markdown, similar to how SoundCast is documented.

Deliverable(s): Frontend routine for preparing land use data inputs with user instructions due six weeks from NTP; Revised Model Setup with Sampling and Stability Improvements along with tech memo documenting cause and effects of results variation of all sub choice models and improvements made, Due 10 weeks from NTP; Revised Model Setup with Simplifications and New Zones and Networks, Due 16 weeks from NTP; Revised Model Setup with Completed PSRC Improvements along with tech memo documenting the revisions, Due 24 weeks from NTP

TASK 4: MODEL CALIBRATION AND VALIDATION

Consultant will calibrate and validate the revised model system to local conditions. Consultant will ensure that the model results in the BKR region are consistent with the guidelines set in TMIP’s Travel Model Validation and Reasonableness Checking Manual Second Edition and that the results outside the region are reasonable as well.
To begin with, Consultant will review the existing model calibration and validation data sets. These include the 2014 household travel survey, the various Sound Transit on-board surveys available, the latest traffic counts and transit line boardings, as well as the Census 2000 Transportation Planning Package (CTPP) and American Community Survey (ACS) 1, 3 and 5 year averages. These data sets will be utilized for this effort.

To facilitate model calibration and application, the Consultant will develop and program summary routines to extract various model statistics by predefined geographic areas such as city, CBDs, Mobility Management Area, and screenlines. A properly validated model ensures that it replicates current travel patterns and responds appropriately to its inputs. Consultant will initially perform a loose calibration of the model system once all the model improvements are in place, and to compare to transit boarding and traffic count data in order to identify aspects of the model that need to be focused on for further attention. This way, we will better understand how the model performs early in the task in order to concentrate our efforts on calibrating and validating key model components as the project progresses. Consultant will meet on-site with the BKR model development team to review and finalize the calibration/validation effort.

Deliverable(s): One-day Onsite Workshop, Revised Model Setup with Final Calibration and Validation; Calibration and Validation Memo; Due 36 weeks from NTP

TASK 5: MODEL DOCUMENTATION AND STAFF TRAINING

Consultant will develop a final BKR model development report by combining all technical memos and document all scripts and datasets used to calibrate/validate the model. Consultant will also develop a User’s Guide that documents steps to:

• Setup a model run
• Execute a model run
• Split zones and change land use inputs for one or multiple zones
• Update and revise model parameters and coefficients

City staff will work closely with Consultant throughout the model development process to assist and learn. Consultant will meet with project team on-site for one day to finalize knowledge transfer in the form of a training workshop.

Deliverable(s): Final Model Development Report, Final Online Model User’s Guide, One-day Onsite Training Workshop; Due 40 weeks from NTP