Calibration and Validation of an Activity-Based Model (ABM)-The San Diego Experience
Corresponding Author: Wu Sun, San Diego Association of Governments
Presented By: Clint Daniels, San Diego Association of Governments
Abstract
This paper describes the calibration and validation of the San Diego Association of Governments (SANDAG) Activity-Based Model (ABM) for base year 2012. Prior to the work described in this paper, there were a few rounds of calibration and validation efforts. This round of calibration and validation is deemed necessary based on the following reasons. First, as demographic and socioeconomic conditions change in the San Diego region, land use and population inputs to ABM also change. Secondly, recent releases of American Community Survey (ACS) make it possible to calibrate some ABM components to represent updated base year 2012 conditions. Previous calibrations relied on outdated 2006 San Diego household travel behavior (HHTS) and Census data. Thirdly, staff added, validated, and cleaned 2012 traffic counts, resulting in improved counts in terms of both quality and quantity.
After evaluating the impact of changes on each model component, the calibration was decided to focus on the following:
•Auto ownership model
•Coordinated daily activity pattern model
•Tour and trip mode choice models
•Crossborder model
•Military trips calibrated against military base gate counts
Roadway validations are by corridors, roadway facility types, directions, and key count locations. Transit validations are by regional transit ridership, transit line haul modes, and transit access modes.
This work resulted in the following conclusions:
•Synthetic population, land use and network inputs, and software changes have significant impact on model results. Model calibration and validation is deemed necessary when these inputs change.
•The fidelity of network coding and traffic counts is important to model calibration and validation, especially for localized model validation.
•Although recent ACS releases provide some updated calibration targets, most calibration targets are still only available in the 2006 HHTS. Fortunately, the San Diego 2016 HHTS is underway and will provide up-to-date calibration and validation targets.
•The calibrated model matches observed counts better than the model pre calibration. This effort helps establish model credibility and confidence of ABM applications in regional planning.