Future climate change may bring risk or benefit to crop production. In this paper, the possible impact of climate change on faba bean production in Scotland is examined. Instead of conventional simulation modelling techniques, the belief network approach is applied to deal with the uncertain information associated with climate prediction. A belief network is constructed which relates faba bean growth and development to weather according to an agricultural expert's advice. The existing FABEAN model and a weather generator are used to generate data to train the belief network, so that it can answer queries on faba bean production under various climate predictions. Using a range of possible future climate scenarios, we have shown that the predictions given by the belief networks can represent accurately those from the exhaustive runs of a simulation model, and in a more efficient way. The possibility of combining belief network techniques with Geographical Information Systems (GIS) as a means of scaling from local to regional predictions for crop production is also discussed.