To explore the role of cloud microphysics in a large dataset of precipitating clouds, a sixmonth dataset of satellite-derived cloud-top brightness temperatures from GOES longwave infrared (channel 4) satellite data over precipitating surface observing stations is constructed, producing 144 738 observations of snow, rain, freezing rain, and sleet. The distributions of cloud-top brightness temperatures were constructed for each precipitation type, as well as light, moderate and heavy snow and rain. The light-snow distribution has a maximum at -16°C, whereas the moderate and heavy snow distributions have a bimodal distribution around -16° to -23°C and a secondary maximum at -35° to -45°C. The light, moderate, and heavy rain, as well as the freezing rain and sleet, distributions are also bimodal with roughly the same temperature maxima, although the colder mode dominates. The colder of the bimodal peaks trends to lower temperatures with increasing rainfall intensity: -45°C for light rain, -47°C for moderate rain, and -50°C for heavy rain. Like the distributions for snow, the colder bimodal peak increases in amplitude relative to the warmer bimodal peak at heavier rainfall intensities. The steep slope in the snow distribution for cloud-top brightness temperatures warmer than -15°C is due to the combined effect of the activation of ice nuclei and the maximum growth rate for ice crystals at temperatures near -15°C. In contrast, the rain distributions have a gentle slope toward higher cloud-top brightness temperatures (-5° to 0°C) due to the warm-rain process. Finally, satellitederived cloud-top brightness temperatures are compared to coincident radiosonde-derived cloudtop temperatures. Although most difference between these two are small amplitude, some are as large as +/-60°C. The cause of these differences remains unclear, and several hypotheses are offered.3

Satellite analysts at the Satellite Services Division (SSD) of the National Environmental, Satellite, Data, and Information Service (NESDIS) routinely generate 24-h rainfall potential for all tropical systems that are expected to make landfall within 24 to at most 36 h and are of tropical storm or greater strength (Ͼ65 km h Ϫ1 ). These estimates, known as the tropical rainfall potential (TRaP), are generated in an objective manner by taking instantaneous rainfall estimates from passive microwave sensors, advecting this rainfall pattern along the predicted storm track, and accumulating rainfall over the next 24 h.In this study, the TRaPs generated by SSD during the 2002 Atlantic hurricane season have been validated using National Centers for Environmental Prediction (NCEP) stage IV hourly rainfall estimates. An objective validation package was used to generate common statistics such as correlation, bias, root-meansquare error, etc. It was found that by changing the minimum rain-rate threshold, the results could be drastically different. It was determined that a minimum threshold of 25.4 mm day Ϫ1 was appropriate for use with TRaP. By stratifying the data by different criteria, it was discovered that the TRaPs generated using Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) rain rates, with its optimal set of measurement frequencies, improved spatial resolution, and advanced retrieval algorithm, produced the best results. In addition, the best results were found for TRaPs generated for storms that were between 12 and 18 h from landfall. Since the TRaP is highly dependent on the forecast track of the storm, selected TRaPs were rerun using the observed track contained in the NOAA/Tropical Prediction Center (TPC) "best track." Although some TRaPs were not significantly improved by using this best track, significant improvements were realized in some instances. Finally, as a benchmark for the usefulness of TRaP, comparisons were made to Eta Model 24-h precipitation forecasts as well as three climatological maximum rainfall methods. It was apparent that the satellite-based TRaP outperforms the Eta Model in virtually every statistical category, while the climatological methods produced maximum rainfall totals closer to the stage IV maximum amounts when compared with TRaP, although these methods are for storm totals while TRaP is for a 24-h period.