Over the last 15 years, extensive research has been done on the transferability of travel demand models. However, much of this work has been concentrated towards investigating the transferability of disaggregate mode choice models. The transferability of an aggregate total demand model for intercity travel is examined. Model transfer is possible only when a number of preconditions for transferability are satisfied. One of the principal obstacles to the successful transfer of intercity demand models is the inability to overcome the contextual differences between calibration and application. Here, the components of the intercity total demand model are separated into exogenous and intrinsic (contextual) factors. The latter is therafter clarified as being either transferable or nontransferable. It is shown that transfereable attributes can accompany a model during transfer. Nontransferable attributes, on the other hand, will free the model of city or city-pair specific contextual characteristics which should not be transferred to other city pairs. The issues involved in transferring an aggregate model are also investigated. Aggregate data on interdistrict travel by public transportation in Sri Lanka have been used to successfully calibrate a total demand model with a number of transferable and nontransferable attributes that represent both temporal and spatial contextual factors. It is shown that the forecasting ability of this model is far superior to a counterpart model without intrinsic variables. (A0.

Research on demand-model transferability has consistently shown that the updated models perform better than the simple transfer of the original model with the original coefficients. Several methods are available for the updating of parameter estimates during model transfer. The scalar factor method has been extended to specify individual factors for each variable. This method allows the flexibility of removing insignificant variables in transfer; it also permits the grouping of parameters that have to be updated by a common factor. INdividual scalar factors can also be identified for variables that are uniquely affected during transfer. This approach therefore incorporates the strength of both the sample data and the calibration model to its maximum showing that this method gives excellent fit to observed flows when tested for geographical transferability of an aggregate intercity total demand model for public transport in Sri Lanka. It is also shown that the Bayesian method becomes less efficient when sample sizes available for updating become smaller.

TRB’s National Cooperative Highway Research Program (NCHRP) Report 716: Travel Demand Forecasting: Parameters and Techniques provides guidelines on travel demand forecasting procedures and their application for helping to solve common transportation problems.

"TRB's National Cooperative Highway Research Program (NCHRP) Report 735: Long-Distance and Rural Travel Transferable Parameters for Statewide Travel Forecasting Models explores transferable parameters for long-distance and rural trip-making for statewide models. Appendixes G, H, and I are not contained in print or PDF versions of the report but are available online. Appendix G presents a series of rural typology variables considered in stratifying model parameters and benchmarks and identifies the statistical significance of each. Appendix H contains rural trip production rates for several different cross-classification schemes and the trip rates associated with each. Finally, Appendix I provides additional information on auto occupancy rates."--Publisher's description.