AN EXPLANATION OF WHY SYNTHETIC TRANSPORTATION STUDIES WORK (ABRIDGMENT)

Although various validation studies have shown that synthetic urban-transportation-study techniques can achieve acceptable results when compared with conventional traffic assignments and ground counts, two questions have been frequently raised: does this favorable comparison indicate that the trip matrix developed by synthetic procedures reflects actual travel patterns?, and how sensitive is the assignment to the input data? This paper discusses a sensitivity analysis that used data from an urban transportation study in Tyler, Texas to help answer those questions. Four matrices were assigned to the same network and compared--the first three of which were assignment matrices and the last being the existing trip matrix: a stochastic trip matrix constrained only to the number of trips in the study area; a stochastic trip matrix constrained only to the number of trips in the study area; a stochastic trip matrix constrained to total trips, trip-length frequency, and trip ends at each external station; and the fully modeled trip matrix. Analysis of the results from the assignment indicates that, as long as an accurate trip-length frequency is used in generating the trip matrix, the assigned vehicle kilometers of travel will very closely match the counted vehicle kilometers of travel, even when the distribution of zonal trip ends is unrealistic. It was concluded that acceptable assignment results can be expected if the following conditions are satisfied: the correct mean trip length is used, the total number of trips in the study area is correct, and a reasonable approximation of the geographical distribution of trip ends is achieved.