In this study, variations in optimal pump-and-treat (P&T) remediation designs and costs for a contaminated and mass-transfer-limited aquifer are investigated for different hydraulic conductivity (K) heterogeneity conditions with focus on the influence of the correlation length (lambda) of spatially variable K values. Several heterogeneous K fields with diverse lambda values and variances (sigma(2)) are considered. The impact of lambda on optimal remediation design selection is analysed considering different relative locations of low and high K regions. Furthermore, optimal designs obtained for different initial contaminant plume configurations are evaluated. Optimal designs are determined using a simulation-optimization approach. Results show that the locations of low and high K zones within an aquifer, and their respective areas defined through lambda, affect remediation design and cleanup cost noticeably. It is observed that in addition to typical geostatistical parameters (lambda and sigma(2)), better determination of both the spatial distribution of low and high K regions and the initial contaminant mass is critical for better P&T design.