Successful solution to any environmental problem implies working with Knightian uncertainty that explicitly deals with decision making under conditions of unstructured randomness. A 'wild' type of randomness that we will never discern due to its unstable properties makes the assignment of corresponding probabilities impossible. For that reason, the consideration of general economical factors within cost/benefit analysis must fail. So, instead of governmental intervention and a cup and trade scheme, we propose a direct financial market of certificates on environmental improvements. The approach is based on the Principal-Agent method. In this study we analyze the problem of one agent only, and take as example deforestation issues. In the analytical part we give examples how these certificates could be defined. We solve some optimization problems in well established stochastic models. The approximate knowledge which actions are environmentally friendly allows us to design good environmental certificates.