Origin-destination (O-D) synthesis methods deal with the problem of deriving trip O-D patterns from traffic counts. A reliable O-D synthesis model for small area (e.g., central business district (CBD)) applications has great potential to help evaluate alternative transportation system management measures. Among various models reviewed, LINKOD was selected for in-depth evaluation because of its apparent suitability for CBD applications. A 1975 San Jose, California, CBD O-D data set with traffic counts estimated by the microassignment model was used to test the performance of LINKOD. Significant differences were found between the synthesized trip table and the base trip table; nevertheless, when assigned to the network by the microassignment model, both trip tables predicted similar flow patterns. Based on these tests, LINKOD was judged to be an acceptable tool for pragmatic applications in CBDs. An extensive sensitivity analysis of the performance of LINKOD was also made to investigate the effects of different initial target trip tables and incomplete link volume counts. Although LINKOD performed best with data on 100 percent of the turning movements, it was found that with 25 percent coverage (plus all cordon-station volume counts) there existed only a 10 to 20 percent loss in synthetic O-D table accuracy. It was also determined that the geographic pattern of the traffic count data affected the outcome considerably. Because a better CBD data set is indispensable for conducting a more complete validation of O-D synthesis models as well as other traffic models, a comprehensive CBD travel data-collection effort appears warranted. (A)
Abstract