Southern California Association of Governments Observations on Congested Speed Estimation System Metrics Group, Inc.
Background Information • SMG presented a congestion map comparing Model results vs. HICOMP at the RTP TAC Meeting. TAC members requested a follow-up on the discrepancies. • Comparisons were conducted based on: • Automated detection data via PeMS • Caltrans 2000 HICOMP Report • SCAG Year 2000 model validation, P.M. peak loaded network
A few facts about the SCAG Model • AM peak period is 3 hours (6-9) and the PM peak period is 4 hours (3-7) • The model speeds are averages over the entire analysis period • SCAG’s Model was developed from Census and Household Travel Survey data and validated to various sources of travel data, including HPMS and the Caltrans Count Book, etc.
A few facts about PeMS detection data • Loop detectors measure volume and occupancy at specific points • District 7 has more than 1,700 detectors which cover 70-80 percent of the district’s freeway system. • Speeds are calculated using volume and occupancy. They are not measured.
A few facts about HICOMP • Freeway segments are identified as congested by operations staff at the Caltrans districts. • Drivers start collecting speed data when actual speeds fall below 35 miles per hour for 15 minutes or longer. • Drivers continue to collect speeds until speeds are above 35 miles per hour. • Multiple drivers are used for each congested stretch of freeway (generally 15 to 20 minutes apart). • Results are mapped, noting location, time, and duration of congestion. In general, duration of congestion does not encompass the entire peak periods as defined by the model.
Congestion Map: Model and HICOMP (Primarily PeMS Estimation)
District 8 Congestion Map: Model and HICOMP (Primarily Tachometer Run Estimation)
Observation and Analysis • Speed calculations using PeMS detection data in the past under-reported actual speeds by up to 15 percent. Discussions with UC Berkeley PATH suggests that refinements are on-going. • With very few exceptions, the overall comparison of model vs. HICOMP congestion estimation appears reasonable/consistent. • Some HICOMP probe-based results do show congestion, but for a shorter period than the 4 hour model period. • The HICOMP results for District 8 represent only a two day sample, which may not reflect true conditions for the average day.
Conclusions • Comparisons of Model speeds using PeMS detection is pre-mature. • Until PeMS is more official, it is not recommend to use it for calibration of SCAG’s 2004 Model. • In the future, PeMS may offer a potentially useful source for model calibration and validation.