Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN) is a satellite infrared-based algorithm that produces global estimates of rainfall at resolutions of 0.25¿ ¿ 0.25¿ and a half-hour. In this study the model parameters of PERSIANN are routinely adjusted using coincident rainfall derived from the Tropical Rainfall Measurement Mission Microwave Imager (TMI). The impact of such an adjustment on capturing the diurnal variability of rainfall is examined for the Boreal summer of 2002. General evaluations of the PERSIANN rainfall estimates with/without TMI adjustment were conducted using U.S. daily gauge rainfall and nationwide radar network (weather surveillance radar) 1988 Doppler data. The diurnal variability of PERSIANN rainfall estimates with TMI adjustment is improved over those without TMI adjustment. In particular, the amounts of afternoon and morning maximums in rainfall diurnal cycles improved by 14.9% and 26%, respectively, and the original 2--3 hours of time lag in the phase of diurnal cycles improved by 1--2 hours. In addition, the rainfall estimate with TMI adjustment has higher correlation (0.75 versus 0.63) and reduced bias (+8% versus -11%) at monthly 0.25¿ ¿ 0.25¿ resolution than that without TMI adjustment and consistently shows higher correlation (0.62 versus 0.51) and lower bias (+22% versus -30%) at daily 0.25¿ ¿ 0.25¿ scale. This study provides evidence that the TMI, which measures instantaneous rain rates from the TRMM platform flying on a non-Sun-synchronous orbit, enables PERSIANN to capture more realistic diurnal variations of rainfall. This study also reveals the limitation of current satellite rainfall estimation techniques in retrieving the rainfall diurnal features and suggests that further investigation of precipitation generation in different periods of cloud life cycles might help resolve this limitation.