Enabling to track and predict severe storm events having potential of causing extreme precipitation and turbulence is critical for flash flood warning and navigation safety. There have been some efforts in developing short-term forecasting algorithms (nowcasting) using ground-based high resolution radar data. Since the radar coverage is very limited, developing a nowcasting technique using satellite information may extend the coverage limitation of ground radar sensors. Typically most of the satellite-based nowcasting algorithms using GOES-infrared (IR) data apply to large scale severe events such as Mesoscale Convective Systems (MCS). The proposesd new satellite-based object tracking algorithm, however, is capable of tracking MCS events as well as local convective storms. The proposed algorithm named PERCAST (PERsiann-foreCAST) predicts the rate of rainfall in the next 1~4 hours, with 4km spatial resolution using past few time steps to extract storms features including advection field and growth and decay trends. The algorithm is coupled with a precipitation retrieval algorithm using satellite information (e.g. PERSIANN-CCS (Precipitation Estimation from Remote Sensing Information using Artificial Neural Network-Cloud Classification System) for precipitation nowcasting. The proposed algorithm is evaluated over the contiguous U.S during summer 2010. The preliminary results show that the proposed approach significantly improved prediction accuracy of several existing short-term nowcasting algorithms (e.g. NOAA's Warning Decision Support System, Integrated Information; WDSS-II). Considering storm growth and decay trends for the prediction may improve the predictability of convections in term of Probability Of Detection (POD) and False Alarm Ratio (FAR) up to 15%.