A clustering algorithm is used to define the radiative, hydrological, and microphysical properties of precipitating convective events in the equatorial region observed by the Tropical Rainfall Measuring Mission (TRMM) satellite. Storms are separated by surface type, size, and updraft strength, the latter defined by the presence or absence of lightning. SST data and global reanalysis products are used to explore sensitivity to changes in environmental conditions. Small storms are much more numerous than mesoscale convective systems, and account for fairly little of the total rainfall but contribute significantly to reflection of sunlight. Lightning storms rain more heavily, have greater cloud area, extend to higher altitude, and have higher albedos than storms without lightning. Lightning is favored by a steep lower-troposphere lapse rate and moist midlevel humidity. Storms occur more often at SST greater than or equal to 28degreesC and with strong upward 500-mb mean vertical velocity. In general, storms over warmer ocean waters rain more heavily, are larger, and have higher cloud tops, but they do not have noticeably higher albedos than storms over cooler ocean waters. Mesoscale convective system properties are more sensitive to SST. Rain rates and cloud-top heights increase statistically significantly with mean upward motion. Rain rates increase with albedo and cloud-top height over ocean, but over land there are also storms with cloud-top temperatures > -35degreesC whose rain rates decrease with increasing albedo. Both the fraction of available moisture that rains out and the fraction that detrains as ice increase with SST, the former faster than the latter. TRMM ice water paths derived from cloud-resolving models but constrained by observed microwave radiances are only weakly correlated with observed albedo. The results are inconsistent with the "adaptive iris'' hypothesis and suggest feedbacks due primarily to increasing convective cloud cover with warming, but more weakly than predicted by the "thermostat'' hypothesis.
586FDTimes Cited:31Cited References Count:43