The PCE was selected as representative DNAPLs. Three 2-D sandbox experiments were conducted to investigate the effect of source-zone architecture on the PCE by Tween 80flushing. The light transmission method was used to monitor the migration and remediation processes of PCE within the sandboxes, and quantitatively measure PCE saturation distribution. The ratio of volume of PCE ganglia to the volume of pools (GTP) was used to quantify source-zone architecture, and the effect of GTP on the PCE removal was evaluated. Experimental results showed that when PCE migrated in the porous media containing lenses, the percolation paths extended, and PCE ganglia increased. PCE ganglia was preferentially solubilized by micelles due to a larger contact area with Tween 80 solution. While PCE pool on the fine-sand layer was much more difficult to remove due to a smaller contact area and limited solubilization capacity. The larger initial PCE ganglia promoted pool on migration path to rapidly transform into ganglia, and hence increased cumulative PCE removal, indicating that GTP significantly influenced the remediation effectiveness of surfactant flushing. Therefore, detailed source-zone architecture characterization is needed to evaluate the remediation efficiency and dose of flushing agent during in-situ surfactant flushing process.