The passive microwave detection of precipitation over land relies primarily on detection of brightness temperature depressions at higher frequencies due to scattering by ice particles aloft. While the principle is straightforward, the signal-to-noise ratio is often quite poor, especially given large spatial and temporal variations in surface emissivity and other environmental variables. A number of ad hoc combinations of sensor channels have been devised over the years by other authors to enable global over-land precipitation detection by the Special Sensor Microwave/Imager (SSM/I), the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), and the Earth Observing System (EOS) Advanced Microwave Scanning Radiometer (AMSR-E). These fixed global methods inevitably have high detection thresholds owing to their need to avoid false rain signatures over certain problem surface types. We have developed a detection and estimation framework for multichannel microwave imagers that is locally optimized based on the annual background brightness temperature covariance (“noise”) and specified target signature within each geographic grid box. In effect, a linear combination of six channels is objectively determined that is more or less blind to temporal and spatial variations in the background, including coastlines and seasonal effects. While the retrieved signature can be directly calibrated and used as a standalone linear rain rate algorithm, we are pursuing the use the retrieved signatures as a low-dimensional “observables” suitable for database-type Bayesian retrievals, as envisaged for the Global Precipitation Mission (GPM).