Modelling is seen by many scientists as a powerful tool to assess the interaction between variability in the bioeconomic environment of small-scale farmers and their resource management strategies. Two case studies on small-scale farming in Mexican and Brazilian savannas are presented. On one hand, the effect of climate variability on crop yields is highly dependent of the managing techniques used. On the other hand, the variability of prices, resulting in Mexico's case from the establishment of NAFTA and, in the Brazilian case, from the "-plano Real," dramatically increases the uncertainty of the economic results linked to technical choices. Modelling the real conditions of such farmer decisions requires integrating of both biophysical and economical knowledge. This work analyses the characteristics required for such models and ways to integrate them. It appears that the building and scaling up of the models should be driven by results of a study on farming system diversity.