Travel Demand Model Calibration for Areas with Heterogeneous Characteristics

T ravel demand model calibration is ~ usually a very cumbersome process that involves several iterative trial-anderror procedures. The degree of difficulty in calibrating a travel demand model depends on several elements, including the scope of the study, the availability of data, the software used, the desired level of modeling precision, the experience of the modeler, and the demographic and socioeconomic characteristics of the study area. One of the most challenging problems in travel demand modeling is the development of calibrated models for large study areas with mixed urban, suburban, and rural localities. Generally, urban, suburban, and rural localities exhibit different travel characteristics. For example, the likelihood of undertaking trips is usually higher in urbanized areas than in rural areas. Conversely, the likelihood of undertaking long trips is usually higher in rural/suburban areas than in urban areas. In addition, a typical urbanized area usually attracts more trips than it produces. Conversely, a typical rural/suburban area produces more trips than it attracts. This is particularly true for work-related, i.e., peak-hour, trips. The production and attraction characteristics of urban, suburban, and rural localities necessitate the use of a composite model, particularly in the trip generation and distribution stages of the modeling process. Attempting to use a common model for travel demand forecasting for a study area with mixed urban/suburban/rural characteristics usually poses many difficulties in achieving favorable results. An increasing number of travel demand model software packages now offer the transportation modeler the flexibility of developing a composite model that can be used to circumvent problems posed by spatial heterogeneity of urban, suburban, and rural localities. Experience has shown, however, that this important feature of travel demand model software packages is seldom exploited. This article describes the methodology that was successfully used to address problems posed by spatial heterogeneity during the calibration of the RADCO travel forecasting model.