A study on the system calibration of the integrated transportation and land use models.

摘要

The problem addressed in this dissertation is that if the zone-to-zone travel costs used in the location model calibrations were to be significantly different from the travel times or costs which resulted from trip assignments, then there certainly would be reason to consider whether the location model calibration should be redone.; A series of the five experiments is designed to provide the empirical evidence of the following questions: how changes in DRAM parameters result in changes in model solutions, how model system responds to a newly estimated parameters or changes in input data such as the population and land use data and impedance matrix.; The experimental scheme adopted in experiment 1 becomes the basis for the next three experiments: subsequent runs of CALIB-DRAM-NETWRK models.; Then, the different scheme is prepared to examine sensitivity of the system to recalibrated DRAM parameters and input data in experiment 2.; Experiment 3 is done with the Houston data to explore the sensitivity of model system to procedures for dealing with variable DRAM parameters.; Another set of runs was done to examine the response of the system to fixed DRAM parameters in experiment 4.; A fifth experiment was done, in a sense, to consider the trip assignment portion of the model package.; The following conclusions were drawn from the five experiments. The first point was that, according to the mean absolute percentage changes for total households obtained, most of the systems appeared to converge to a stable solution under the different frameworks. The second major point was that once a system appears to achieve its convergent solution, a significantly different result is obtained from the use of a newly estimated set of parameters. With the use of a 80% scaling factor, it was shown that the level of congestion effected the model solution procedures.