Numerical model experiments are conducted to assess how aircraft icing forecasts are affected by the choice of microphysical parameterization, with respect to the presence of supercooled liquid water (SLW). Select winter storms are investigated that are observed to have elevated regions of SLW, a condition for hazardous icing environments. The schemes considered include the Thompson, Milbrandt-Yau, Morrison, and National Severe Storms Laboratory (NSSL) microphysics parameterizations. The Thompson scheme routinely produces less SLW than the other three schemes. Sensitivity experiments reveal this difference is in large part a consequence of scavenging of cloud water from snow. In some environments, the Morrison scheme produces considerably more SLW than the other parameterization schemes. The output from each experiment is compared to aviation pilot reports of aircraft icing to determine which scheme yields the most accurate results. According to all statistical measures that we explored, the predictability of SLW seems to be best represented by Morrison. However, the results are not drastically different between the schemes.