Sound policy needs tools for production analysis under uncertainty such as the state-contingent approach. Apart from demonstrating the potential applicability of the approach for the grain producing sector in Saxony-Anhalt, Germany, this contribution addresses one of the key challenges to its empirical applications in the context of agricultural production analysis under uncertainty, namely the definition of the states of nature and the construction of environmental indicators to detect the occurrence of the so-defined states. This study proposes such an environmental indicator based on biophysical crop yields and suggests two methods for statistical data clustering to evaluate its values. To the extent to which simulated data reflect key characteristics of the real data this contribution presents estimates for the state-contingent production technology. The hypothesis of an output-cubical technology is rejected in contradiction to assumptions implicit to conventional production analysis under uncertainty.