In this study, we develop a marginal chance-constrained data envelopment analysis model in the presence of non-discretionary inputs and hybrid outputs for the first time. We call it a stochastic non-discretionary DEA model (SND-DEA), and it is developed to measure and compare the relative efficiency of forest management units under different environmental management systems. Furthermore, we apply an output-oriented DEA technology to both deterministic and stochastic scenarios. The required data are collected from 24 forest management plans (as decision-making units and included four inputs and equal amount of outputs. The findings of this practical research show that the modified SND-DEA model in different probability levels give us apparently different results compared to the output from pure deterministic models. However, when we calculate the correlation measure, the probability levels give us a strong positive correlation between stochastic and deterministic models. Therefore, approximately 40% of the forest management plans based on the applied SND-DEA model should substantially increase their average efficiency score. As the major conclusion, our developed SND-DEA model is a suitable improvement over previous developed models to discriminate the efficiency and/or the inefficiency of decision-making units to hedge against risk and uncertainty in this type of forest management problems.