The performance of algorithms retrieving cloud liquid water content W from radar reflectivity factor Z is evaluated using data from a three-dimensional large-eddy simulation model with detailed size-resolving microphysics. On the basis of case studies of a marine stratocumulus cloud observed during the Atlantic Stratocumulus Transition Experiment in June of 1992 and a continental stratus cloud observed over north central Oklahoma on April 30, 1994, it is shown that retrieval algorithms based only on Z are very sensitive to cloud microphysics and do not reproduce adequately the horizontally averaged vertical profile of W. Additional independent measurements are needed to constrain the retrieved W profile. Surprisingly, algorithms show little improvement when the exact concentrations of cloud droplets obtained from the model are included in the calculations. Vertically integrated liquid water path P is found to be a robust constraint that ensures a more accurate retrieval. In the two studied cases an algorithm based on the Z-W relation of the form Z=aWb gives an optimal performance when a is inversely proportional to P and b=1.32. The horizontal averaging of P over a computational domain of several kilometers does not affect the accuracy of the retrieval of the horizontally averaged profile of cloud liquid water content, which adds to the versatility of the algorithm. ¿ 2000 American Geophysical Union