Project Meeting 2017.07.20 - bellevuewa/BKRCast GitHub Wiki

Low CBD volumes

  • We created some additional summaries to understand the issue:
    • Assignment volume bars colored by over/under the count for AM and PM
    • Trips by purpose to/from districts by time-of-day
    • Person type and trip purpose by district
    • Work location versus employment by district
  • Based on the assignment maps, Hu fixed unassigned freeway ramps in Bellevue and then we re-ran the model, which helped a bit, but not enough
  • The trips by purpose summary showed not enough work travel
  • We tried increasing the attractiveness of office employment and high accessibility in work location choice and initial runs before shadow pricing looked better, but after shadow pricing results were basically the same
  • Summaries of workers to the CBDs show the number of workers in the synthetic population (synpop) is too low (see below)
  • Work location shadow pricing is correcting the in-balance as best it can so even though we're making Bellevue CBD more attractive, the lack of workers for jobs in other districts means Bellevue CBD is low on workers as well

Workers to employment testing

  • 1.8m workers in the synpop and 2+m in the employment
  • As a quick test, we scaled up hhs with workers and scaled down hhs without workers
  • The moreworkers.py script operates on the PSRC synpop file and takes as input a scaling factor by TAZ
  • It sums the number of persons after the adjustment and scales down non-worker hhs to try and match the previous total number of persons by taz
  • This results in 2m workers, and about 35k more persons in the region
  • The result is the same synpop file but with modified hh expansion factors
  • We then ran this synpop through the model and the results look better:
    • workers to Bellevue CBD match better: 32k before and now 40k, which matches LODES
    • work trips better match the survey (and also for other purposes for worker person types)
    • assignment results for AM and PM versus the counts match better as well
  • We've seen this shortfall in workers in San Diego and Fresno as well
    • Employment has been growing year over year, especially in places like BKR and the ACS 5-year average is low
  • FOLLOW-UP
    • Mark B and Mark S looked at ACS and CPS/LAUS employment data and 2015 employment is 8.2% higher than the 2010-2014 average
    • PSRC agrees that an updated synpop should be built, and will help, but doesn't have the time to do it right now
    • @Bellevue and PSRC to discuss some more

Project management

  • This is our last scheduled meeting
  • @Nagendra update setup (including adding the more workers script) and share with Hu
  • @Hu run latest setup with ESD data, including output summaries
  • @Nagendra to share updated calibration worksheets and scripts
  • We agreed to declare calibration/validation done at this point
  • The last remaining deliverable is a webinar training/support discussion at the end of August, which gives Bellevue some time to get more familiar with the model
  • Bellevue is also looking into options for additional support if desired
  • Items on the list for the next round of development include the updated synpop and PSRC's adjustments for walk trip length