We conduct a functional benefit transfer over 20 years of total willingness to pay based on car-borne forest recreation in 52 forests, using a mixed logit specification of a random utility model and geographic information systems to allow heterogeneous preferences across the population and for heterogeneity over space. Results show that preferences for some forest attributes, such as species diversity and age, as well as transport mode have changed significantly over the period. Updating the transfer model with present total demand for recreation improves the error margins by an average of 282%. Average errors of the best transfer model remain 25%.