This paper examines using cellphone Origin-Destination (OD) data purchased from AirSage, Inc. for validating a recently updated version of the Syracuse Metropolitan Transportation Council (SMTC) Regional Travel Model. The cellphone OD data is essentially a zone-to-zone origin-destination matrix containing the number of trip counts occurring between each OD pair. The study area for this analysis had 800 zones and 2 million daily regional trip counts disaggregated by time of day (AM peak, Mid-day, PM peak, and 24 hour totals), imputed trip purpose (Home-based work, Home-based other, and Non-home based) and cellphone user classification (resident or visitor).
The cellphone-derived OD matrix was compared against the OD matrices estimated by the regional model. A number of performance indicators were applied:
- Total trips by trip purpose
- Total trips by purpose for important regional generators (eg. University of Syracuse)
- Select subarea and district level trips flows
- Average trip lengths and trip length frequency distributions
- Time of day distributions
- External Trip Flows (IX, XI and XX)
The comparison suggests that the fidelity of the cellphone OD data diminishes with disaggregation – spatially, temporally and by purpose. Users should be mindful of performing direct comparisons of cellphone OD data against modeled output because a cellphone movement is usually not identical to a modeled trip. Some major reasons for the differences include:
- Cellphone OD movements are not “linked” trips as they are in the modeling context;
- Cellphone locations are generally collected only when the device interacts with the cellular network;
- OD locations and purposes are imputed based on statistical analysis; and
- Some cell movements are not germane to the scale of regional modeling.
The use of cellphone data is gaining traction in the industry because it offers a number of important and exciting advantages. The presentation will describe how cellular data products are and can be useful in this context but not as a wholesale replacement of traditional methods and data collection activities. The data collected via emerging technologies and the use of ‘Big Data’ should be used in conjunction with other methods, data sources, and professional judgment when informing a model calibration and validation effort.