The success of cleaning up a contaminated site depends on the degree of knowledge of the site's characteristics. When there is much uncertainty associated with the knowledge, the uncertainty regarding whether the remediation will work increases consequently. It is therefore essential to know how much reduction of uncertainty is needed for the purpose of designing a successful and reliable remediation system. The understanding of the site characteristics is basically increased by site investigation, and thus the uncertainty is decreased by sampling information. This study develops a method to evaluate the value of reducing uncertainty by sampling the hydrogeological parameters in a groundwater remediation system. Hydraulic conductivity being taken as an example of the site characteristics, random field generation and conditional simulation are coupled to obtain a range of hydraulic conductivity fields based on the sampling outputs. A multiple-realization management model that incorporates a chance constraint of health risk is used to find the lowest remediation cost under specific remediation criteria of risk through genetic algorithm. The remediation cost, which is expected to decrease with collection of more samples, serves as the measure of the value of uncertainty reduction by sampling. A case study shows that the variation of the hydraulic conductivity fields among the potential sites as well as the remediation cost is reduced as a result of increase of samples. It also shows that the risk after remediation decreases with the collection of more samples, which implies that the reduction of risk can also be used to assess the value of sampling.