The capacity of tourism destinations to attract visits and the propensity to make round trips to remote sites were approached through a model based on populations, travel times, traffic on road links, and identification of prime tourism destinations. A genetic algorithm was used to simultaneously estimate a gravity model of trip generation, incorporating an attraction population multiplier, and a route assignment model. Local residents and tourists were covered in separate but additive parts of the model. Results indicate an attraction multiplier of approximately four and a high propensity for long-distance car tourists to return home by a different route. (C) 1997 Elsevier Science Ltd.