The mass removal behavior of pooled light nonaqueous phase liquid (LNAPL) by air sparging was investigated using one-dimensional and two-dimensional physical model experiments. The effects of flow rate, flow interruption, nonaqueous phase liquid (NAPL) saturation, and soil heterogeneity were examined. At the beginning of air sparging, a higher initial mass removal rate was achieved under near-equilibrium condition but was limited to the volatilization of NAPL that is in direct contact with air channels. At the later stage of air injection, the mass partitioning from NAPL phase to gaseous phase became rate limited. In terms of final amount of mass removed, there was a negligible effect of air injection flow rate within the range examined. Results from two-dimensional soil tank experiments revealed that fractional mass removal was extremely sensitive to subsurface heterogeneity. About 90% fractional mass removal was achieved in a case where NAPL was trapped in a coarse sand lens surrounded by a finer soil matrix. However, when NAPL was trapped in a fine sand lens in coarser sand matrix, the injected air channels diverted away from the entrapped NAPL due to the presence of the low-permeability sand lens, and the fractional mass removal decreased to less than 50%. Flow interruption can be used to enhance the mass removal of LNAPLs entrapped around the water table by air sparging and soil vapor extraction. However, the fractional mass removal rate became zero at a threshold NAPL saturation beyond which NAPL could not be removed by air sparging even after many interruption cycles. Results indicate the necessity of incorporating a rate-limited mass transfer model into the NAPL-gas mass transfer model. The mass transfer rate should be a function of NAPL saturation and the threshold value determined by experiments.