The influence of microwave radiometer accuracy on retrieved cloud liquid water path (LWP) was investigated. Sensor accuracy was assumed to be the sum of the relative (i.e., Gaussian noise) and the absolute accuracies of brightness temperatures. When statistical algorithms are developed the assumed noise should be as close as possible to the real measurements in order to avoid artifacts in the retrieved LWP distribution. Typical offset errors of 1 K in brightness temperatures can produce mean LWP errors of more than 30 g m-2 for a two-channel radiometer retrieval, although positively correlated brightness temperature offsets in both channels reduce this error to 16 g m-2. Large improvements in LWP retrieval accuracy of about 50% can be achieved by adding a 90-GHz channel to the two-channel retrieval. The inclusion of additional measurements, like cloud base height from a lidar ceilometer and cloud base temperature from an infrared radiometer, is invaluable in detecting cloud free scenes allowing an indirect evaluation of LWP accuracy in clear sky cases. This method was used to evaluate LWP retrieval algorithms based on different gas absorption models. Using two months of measurements, the Liebe 93 model provided the best results when the 90-GHz channel was incorporated into the standard two-channel retrievals.