Temporal information from a geostationary (GEO) satellite is explored for improved precipitation characterization and quantification. Temporal predictors are extracted from the Geostationary Operational Environmental Satellite (GOES)-16 Advanced Baseline Imager (ABI). A 1-D convolutional neural network (CNN) architecture is proposed and compared to a deep neural network (DNN) benchmark without temporal predictors. While the CNN detection (rain/no-rain separation) is on par with the DNN benchmark, promising results are obtained for both precipitation type classification and quantification, indicating the value of the temporal information for precipitation estimation. The identification of rapidly evolving convective systems is improved (accuracy improved from 49% to 58%), while false detections in stratiform-type precipitation are reduced. Rain rate quantification is improved by reducing the overestimation of low rain rates, which increases the correlation from 0.34 (DNN) to 0.41 (CNN). This study is an important first step toward better characterizing and quantifying precipitation using the very high temporal resolution of GEO satellites. © 2004-2012 IEEE.