Project Meeting 2016.08.18 - bellevuewa/BKRCast GitHub Wiki
Employment Data
- Washington State employment department will not grant consultants access to the ES202 (QCEW) data that was used to develop SoundCast
- This is uncommon since just about every travel model in the country uses this data
- Often a confidentiality agreement must first be signed
- This is an issue since it means we cannot run the model and do calibration/validation
- BKR and PSRC are going to talk with the employment department to see if this can change
- RSG to provide examples of confidentiality agreements from other states
- RSG to use a fake data set for the time being
PSRC Cooperation
- PSRC plans to revise the supplemental models to operate at the parcel level this Fall so they can more easily work with the new BKR model
- PSRC plans to complete limited model re-estimation in October and calibration/validation February.
- PSRC will give us an interim release in November/December so we can start our calibration/validation exercise with the latest and greatest
- PSRC will help validate the bike route choice model this Fall; BKR to build bike count database and attach it to the network
- SoundCast model run summaries scripts already on GitHub
Model Design
- We will use BKR’s new zones for auto and transit skimming and assignment
- We will use parcels as the key spatial unit for Daysim
- We will OpenStreetMap and likely either EMME or pandanas for short distance shortest path calculations
- We want to use EMME bike network skims as much as possible since they include generalized cost (and not just distance for the allstreets procedures)
- PSRC not doing toll/non-toll but VOT classes instead so BKR will just do this as well
- New car sharing mode will focus on Uber/Lift/Taxi and not so much on AV; so HOV skims
- Can we allow for intra-parcel trips? This is important for mixed use on big parcels – UW, Microsoft, etc.
- 2014 will be the model base year
Land Use Data Development Process
- We agreed that BKR will interface with the model at the zone level
- There will be a tool that disaggregates zonal level data – hhs, employment, etc. – to the parcel level
- Need to describe / account for splitting of zones
- The tool doesn’t do forecasting; just format conversion
- PSRC to think about how best to assist BKR and others with this in the future
Randomness Testing
- Will merge the various randomness tests and come up with a systemic plan for ~ 10 tests or so
- Spatial population sampling plan will likely look something like: BKR 500%, Seattle 25%, remaining PSRC 10%
- Random seeding is synchronized across runs by resident/tour/trip/model combination so this helps with stability
- Will review Bellevue square and Microsoft as well since they are large trip generators; and also during calibration/validation
Next Steps
- RSG work on a full model run with the new BKR zones and network and quick-and-dirty revised model inputs
- RSG finalize model design and randomness testing plan in the form of a technical memo
- PSRC share full model setup, including inputs (with random employment)
- Everyone work on the employment issue