An evaluation of system analysis techniques for obtaining short-term quantitative precipitation forecasting, based on radar images, is presented. The analysis considers eight rainfall events monitored through radar scans collected at San Pietro Capofiume (northern Italy). These images have temporal and spatial resolutions equal to 15 min and 1 ¿ 1 km2, respectively, and are composed by 256 ¿ 256 pixels. The nowcasting is performed by considering four different approaches, ranging from purely heuristic techniques to system analysis methods for modeling storm advection and dynamic storm evolution in a Lagrangian reference frame, traveling along with the storm. Stochastic models of the autoregressive type, as well as neural networks, are applied for modeling the Lagrangian storm dynamics. The sensitivity of the forecasting performances with respect to prediction lead time and spatial resolution of the radar images is investigated. Finally, the study presents a meta-Gaussian approach for the quantitative evaluation of the uncertainty of the forecast. This method aims to obtain an estimate of the probability distribution of the nowcasting error, conditioned on the predicted future rainfall depth.