Project Meeting 2016.08.25 - bellevuewa/BKRCast GitHub Wiki

Employment data

  • State attorney general interpretation that only employment department consultants can access ES202/QCEW data
  • Project team will continue to work on changing this
  • RSG to share other state employment data confidentiality / NDA forms
  • PSRC to compare LEHD employment data model runs to current model runs
  • Maybe mashup LEHD, QCEW, InfoUSA instead so source of data unknown?
  • Project will need best employment data at the parcel level by Nov/Dec

Full model run

  • RSG started working on getting the quick-and-dirty BKRCast model setup
  • Using new zone system and network
  • PSRC shared latest model setup and inputs (except for parcel employment) with RSG
  • BKR still refining network so will need to batch in again at a later point
  • Will share with BKR once working

Sampling plans

  • Mark discussed draft plan for evaluating sampling, randomness variation
  • Three stages – spatial sampling, destination choice sampling, overall model feedback sampling
  • Spatial sampling
    • 5 spatial sampling options and 10 random seeds
    • Therefore 50 1 iteration Daysim runs; no EMME stuff at this point
    • Review standard deviation of outputs for common metrics
    • Summarize at the region, BKR, district, and TAZ level
  • Select 2 most promising spatial sampling plans and do different destination choice sampling options
    • 2 by 10 seeds by 3 dest choice samples = 60 model runs
  • Select 2 most promising dest choice sampling plans and do different overall model feedback sampling options
    • 2 by 10 seeds by 3 samples = 60 model runs
    • Will run EMME and include assignment metrics as well
  • Hu to draft new district scheme based on BKR neighborhoods
  • RSG started implementing spatial sampling plan in Python
    • Reads hdf5 full population file, TAZ sample rate file and other settings, and writes new sampled hdf5