Estimating Network Model Parameters from Mail Survey Data

by Norman L. Marshall, Resource Systems Group, Inc, Norwich, United States,
Kenneth H. Kaliski, Resource Systems Group, Inc, Norwich, United States,
Leslie L. Rimmer, Resource Systems Group, Inc, Norwich, United States,
Stephen J. C. Lawe, Resource Systems Group, Inc, Norwich, United States,

Document Type: Proceeding Paper

Part of: Microcomputers in Transportation


Network models are generally based on a four-step modeling process; trip generation, trip distribution, modal split, and trip assignment. Microcomputer implementations of these steps have become readily available. However, there are often inadequate resources to fully develop all four steps. One common short-cut is to use default trip generation and trip distribution parameters. This can lead to significant model inaccuracies. Furthermore, trial-and-error adjustments of these coefficients in order to improve calibration can be extremely time consuming, and may not result in an accurate forecasting model. A set of techniques for estimating these parameters from mail surveys was developed and applied to the Burlington, Vermont region, the City of Keene, New Hampshire, and the Syracuse, New York region. In each area, household trip diaries were collected for a sample of 400 to 1000 households. The survey data were used to estimate production rates based on residential land use, attraction rates based on non-residential land use, and trip distribution (gravity model) parameters. The resulting parameters differ considerably from area to area, and differ even more from published default values. Well calibrated models were achieved using the estimated parameters.

Subject Headings: Trip distribution | Traffic models | Parameters (statistics) | Traffic surveys | Forecasting | Errors (statistics) | Urban and regional development | United States | Vermont | New Hampshire | New York

Services: Buy this book/Buy this article


Return to search