A mixed-integer linear programming model for optimisation in the roundwood supply chain has been developed. Many considerations of importance to this chain cannot be predicted without errors. In such cases, it is often optimal to make flexible investments in the initial period and later on sequentially and optimally adapt decisions to the true and observed outcomes of the stochastic variables. This is called adaptive optimisation. The roundwood supply chain is critically dependent on roads. Access to roads strongly depends on the current and previous levels of rain and snow. Therefore, it is natural to use adaptive optimisation when trying to optimise the roundwood supply chain. The model structure developed is explained by a sample problem. The structure can be filled with spatial information from the road network in any area. A general adaptive optimisation model is presented here, and tested with an example. The results indicate that deterministic models cannot be used to model the roundwood supply chain correctly.