In this paper the theory and results of a trip-distribution model that uses a multimodal composite definition of impedance as its measure of separation, instead of highway time, are presented. The distribution model is part of a complete travel-demand model chain developed for the New Orleans region. This model chain is briefly described, and its special features of income stratification and connectivity among programs are emphasized. The disutility functions of a three-mode logit modal-choice model are used to develop modal impedance values. The structure and coefficients of these equations are discussed. Two alternative methods for combining these modal impedances are presented: harmonic mean and log sum. A special technique for calibrating the F factor curves was developed to circumvent shortcomings in the urban transportation planning system (UTPS) software. The results of the calibration are presented. These results indicated that the log sum formula produced better results than the harmonic mean formula, based on various observed and estimated comparisons. In addition, the log sum composite impedance-based model proved suitable only for home-based work trips. Unsatisfactory results for the other trip purposes led to the use of off-peak highway time for those purposes. Results for home-based other and non-home-based models are also presented. The conclusions of this analsysis are that a distribution model can be successfully calibrated by using composite impedance; that, at least in this case, the log sum formula worked better than harmonic mean; and that a successful alternative to the standard AGM gravity model calibration process can be developed.
Abstract