This study involves the extensive evaluation of six years (January 1998--December 2003) of the research-version TMPA three-hourly 0.250 W 0.250 rainfall product over a pixel containing 23 rain gauges from the Micronet network in Oklahoma, United States. The network average rainfall is characterized by small (2%) error. The authors consider three seasons i.e. cold, warm, and hot, and the evaluation is based on scatterplots, comparison of the tail of the distributions, and different types of statistical measures of agreement, bias, disagreement, and categorical evaluation. Overall, the best agreement is for the hot season. This study reveals that the TMPA should be considered as a 100-minute accumulation product, with the accumulation starting between 90 and 30 minutes before its nominal time, depending on the season.