We examined the potential use of standard optimization algorithms as implemented in the inverse modeling code iTOUGH2 (Finsterle, 1999abc) for the solution of aquifer remediation problems. Costs for the removal of dissolved or free-phase contaminants depend on aquifer properties, the chosen remediation technology, and operational parameters (such as number of wells drilled and pumping rates). A cost function must be formulated that may include actual costs and hypothetical penalty costs for incomplete cleanup; the total cost function is therefore a measure of the overall effectiveness and efficiency of the proposed remediation scenario. The cost function is then minimized by automatically adjusting certain decision or operational parameters. We evaluate the impact of these operational parameters on remediation using a three-phase, three-component flow and transport simulator, which is linked to nonlinear optimization routines. We demonstrate that the methods developed for automatic model calibration are capable of minimizing arbitrary cost functions. An example of co-injection of air and steam makes evident the need for coupling optimization routines with an accurate state-of-the-art process simulator. Simplified models are likely to miss significant system behaviors such as increased downward mobilization due to recondensation of contaminants during steam flooding, which can be partly suppressed by the co-injection of air.