The development of a satellite infrared technique for estimating convective and stratiform rainfall and its application in studying the diurnal variability of rainfall in Amazonia are presented. The convective-stratiform technique (CST), calibrated by coincident, physically retrieved rain rates from the TRMM microwave imager (TMI), is applied during January--April 1999 over northern South America. The diurnal cycle of rainfall, as well as the division between convective and stratiform rainfall, is presented. Results compare well (a 1-hour lag) with the diurnal cycle derived from TOGA radar-estimated rainfall in Rondonia. The satellite estimates reveal that the convective rain constitutes, in the mean, 24% of the rain area while accounting for 67% of the rain volume. Estimates of the diurnal cycle for the entire Amazon Basin are in agreement with those from the TRMM precipitation radar, despite the limited sampling of the latter. The effects of geography (rivers, lakes, coasts) and topography on the diurnal cycle of convection are examined. In particular, the Amazon River, downstream of Manaus, is shown to both enhance early morning rainfall and inhibit afternoon convection. Monthly estimates from this technique, dubbed CST/TMI, are verified over a dense rain gage network in the state of Ceara, in northeast Brazil. The CST/TMI showed a high bias equal to +33% of the gage mean, indicating that possibly the TMI estimates alone are also high. The root-mean-square difference (after removal of the bias) equaled 36.6% of the gage mean. The correlation coefficient was 0.77 based on 72 station months.