Bio-economic farm models have been widely used in agricultural research for ex-ante analysis and evaluation of policy decisions. The main goal is to simulate the response of a farm or the behavior of the farmer after policy reorientation and change. Since models are by definition representation of reality, it is expected that the simulated farmer's behavior will partially differ from what is finally observed in reality. This research project focuses on the exploration of potential ways to improve the performance of bio- economic farm models which are used to predict farmer's behavior. Important elements of reality that need to be included in the model in order to eventually simulate the actual farmer's behavior will be indicated, while calibration of model relations and parameters will be considered. The developed methodology will be implemented in an existing bio-economic farm model. It is important to mention that this project is part of a large EU project called SEAMLESS.