The scattering signature in various microwave frequencies from atmospheric ice particles was directly related to surface rainfall rate. Recently, Zhao and Weng (2001) developed a new algorithm to simultaneously derive cloud ice water path (IWP) and ice particle effective diameters (De) using latest sensors AMSU-B measurements at 89 and 150 GHz. It was found that the AMSU derived ice cloud products are strong-correlated with surface rain rate in many storm cases. In this study, a physical approach is proposed to quantitatively relate the rainfall rate to ice water amount and particle size that are derived from AMSU-B. The algorithm consists of a screening method to separate non-precipitating ice clouds from precipitating ones, and an IWP and rainfall rate relationship that is developed using cloud model results. It is found that the AMSU-B measurements at 150 GHz are very sensitive to light precipitation due to the increasing sensitivity of scattering to smaller ice particle sizes in the rain layer. The retrieved rainfall rate are compared with that derived from radar and gauge data. Monthly mean precipitation products from AMSU are also compared with that from SSM/I. It is shown that the algorithm accurately depicts both localized rainfall events and rainfall patterns on the large scale. For a more accurate retrieval of rain rate, it is necessary to group ice clouds into several classes so that separate IWP ~ rain rate relationships can be derived and used for different type of rain scenarios, such as, stratiform and convective type. Further improvement of the algorithm is ongoing toward this direction.