Impact of Conditioning on Robust Remediation Design

Uncertainties in input parameters, such as hydraulic conductivities may result in inconsistencies between simulated and observed groundwater flow and transport data. If a groundwater management model for aquifer remediation design is based upon incorrectly estimated flow and transport parameter data, non-optimal remediation policies may result. To improve upon deterministic management, which is based on a single calibrated input parameter set, previous studies have developed robust optimization models for remediation design. Robust optimization searches for a remediation policy that is likely to perform well over the range of conditions that might be observed in the field. The range of conditions is simulated through generation of multiple realizations of unconditional random fields of the uncertain parameters. These robust optimization studies have explored the relative impact of the degree of uncertainty in various input parameters on the robust management decisions. This study expands upon previous work through incorporation of random fields conditioned on localized field observations. The importance of conditioning on both the final remediation design and on the conclusions about the overall impact of parameter uncertainty is explored by comparison to robust optimization using unconditional random fields and by varying the number of data points utilized for conditioning.