In combination with crop growth models, farm-level models allow an in-depth, process-based analysis of farmer adaptation to climate change and agricultural policy. Evaluated for all farms in an area and extended by interactions, farm-level models become agent-based models that allow simulating aggregate regional production and structural change. Confined to a local or regional scope, however, they cannot directly incorporate price feedbacks that play out at global scale. In this contribution, we use experimental designs to evaluate a non-connected agent-based model for the full space of potential future price developments. We discuss and compare the use of standard regression analysis and non-parametric, automatic methods (MARS and Kriging) to summarize supply behavior over the simulated price ranges. Estimated supply functions constitute a surrogate model for the original agent-based model and could be used to iterate detailed regional analysis with national or global market models in an efficient way.