A new precipitation retrieval scheme for use in the tropics is developed using a combination of visible (VIS)/IR and microwave satellite observations. The new scheme combines the advantages of both the ample coverage of VIS/IR sampling and the physical directness between precipitation and microwave radiances. A VIS/IR algorithm is ''trained'' using Special Sensor Microwave/Imager (SSM/I) derived values of precipitation. Cloud top temperatures and visible optical depths provided by the International Satellite Cloud Climatology Project (ISCCP) level DX analyses are ''binned'' by every 10 K and every 10 unit optical depth, respectively. Lookup tables for the probability of rain and mean rainfall rate are constructed for each cloud top temperature/optical depth cell whenever the optical depth is available (daytime). For the nighttime, the tables are based only on cloud top temperature. The instantaneous rainfall rates are obtained by multiplying the mean rainfall rate by the probability of rain, both characterized by the cloud top temperature and/or visible optical depth associated with the satellite pixel. The satellite retrieval always retains results from optimal parameters, which means SSM/I results are used whenever SSM/I samplings are available; otherwise, VIS/IR results are used during daytime, while IR-only results are used during nighttime. The final precipitation product can have resolution as high as 3 hours in time and about 0.3¿ in space. However, averaging in space and/or time improves the statistics of the derived rainfall rates, since the statistics of the algorithm are more fully utilized. The satellite-derived rainfall product compares well with ship rain gauge data in terms of rain/no rain determination, and agrees well with daily rainfall accumulation derived from a large-scale moisture budget. The satellite retrieved rainfall rates are biased high relative to radar observations. In the absence of satisfactory data with which to validate the satellite-derived rainfall rates, it can be concluded that there is useful information in the ISCCP data set that relates to precipitation and that the inclusion of visible optical depth information improves the daytime retrieval of precipitation relative to IR-only methods. ¿ American Geophysical Union 1996