Optimization using the EnOpt method discussed in the previous chapter can be expensive for realistic field models. We are therefore encouraged to look for less computationally costly alternatives that reduce the number of required reservoir simulations. One such approach is to optimize the controls using the mean reservoir model, as was discussed in Sect.  10.4 , rather than optimizing over the whole ensemble of geological models. When the geologic uncertainty is small, or the modeled response to changes in the controls is weakly nonlinear, this approach may give reasonable results. However, in many realistic cases, optimizing the objective function over the whole model ensemble will provide a different result than the solution obtained when we optimize using the mean model. This chapter will delve into an approach designed to reduce the bias in the approximate objective function, which is a crucial step in ensuring the accuracy of our optimization process.