One key precipitation objective is repeated continental or global observations at the time and spatial scales of convective precipitation, e.g., ~15 km at ~15 minutes.  This is practical only from satellites in geostationary or comparable orbits.  The attached PowerPoint presentation, to be delivered at IGARSS in Denver on July 31, 2006, compares precipitation retrieval accuracies for alternative passive microwave imaging spectrometers compatible with GOES and METEOSAT use.  The paper is "Precipitation retrieval accuracies for geo-microwave sounders" by D. H. Staelin and C. Surussavadee.

The instruments evaluated in the presentation include parabolic Cassegrain antennas and aperture synthesis systems, where the former use frequencies 118-425 GHz to obtain good spatial resolution with reasonable antenna sizes (1.2-2 m diameter), and the latter use ~600-900 microwave receivers at ~50-190 GHz for the same reason.  The retrieval algorithm architecture is the same for all systems and employs multiple neural networks trained using cloud-resolving MM5/TBSCAT simulated radiances at 5-km resolution, smoothed to the appropriate fields of view.  The Goddard explicit cloud model yields profiles of snow, graupel, cloud ice, and rain water.  Several of the slides trigger movies that are not included, but should be available at the Melbourne meeting of the IPWG in October 2006.

The two principal conclusions from this study are that: (a) geostationary microwave spectrometers are indeed capable of providing useful precipitation data with the desired 15-km/15-minute resolution, and (b) a 1.2-meter parabola can yield surface-precipitation-rate retrievals with utility comparable to that expected from a 900-receiver aperture synthesis system observing 8 channels at 54/183 GHz, and superior to those produced by a 600-receiver 8-channel 54/118 GHz system.  The weight, power consumption, risk, and cost of a 1.2-meter antenna with only four receivers would appear to be much less than for systems with 600+ receivers and antennas.