Small-scale, high-impact weather events pose a severe threat to the humans, such as fast- evolving supercells and flash floods. Forecasts of these small-scale storms with a short lead time are of significant value for better preparation and response. Currently several nowcasting algorithms exist to provide the short-term forecast for tens of minutes up to a few hours. These nowcasting techniques can be classified into two categories: object-oriented approaches and correlation-based approaches. In the former technique, abstraction of the radar measurement based on centroid analysis is employed to detect major storm features from radar data where a storm cell is modeled as mass-weighted centroids. Nowcasting is realized as tracking such high-level abstractions. This technique has been shown to work well in predicting motion of well-organized, isolated cells of high reflectivity, but essentially this object-oriented description lacks the capability to characterize the space/time variability of the storms. In the latter technique, local area correlation is computed over subsequent radar observation without relying on any abstraction. The distribution of correlation characterizes the storm evolution and can be used to estimate storm motion. Optimal performance can be achieved only with a window size and orientation matched with the specific storm characteristics at cost of relatively high computational complexity. Recently, a new nowcasting algorithm, the Dynamic and Adaptive Radar Tracking of Storms (DARTS), was developed over a physical model that is described as a flow equation of the continuous spatiotemporal reflectivity field. The governing flow equation is solved in the spectral domain to estimate the storm motion. In this process, correlation computation is avoided and the storm motion can be efficiently estimated with less computational demand. Various scales of storm motion prediction can be controlled by the choice of the number of spectral coefficients used in the estimate. The estimated motion field can be globally constructed over the whole spatial region where radar images are rendered so that the storm observation can be continuously advected forward. Therefore, the DARTS algorithm exhibits several advantages of particular operational importance over other nowcast methods, especially in the real-time operational environment. This paper presents the development of a real-time DARTS nowcasting system in a networked radar testbed operated by the center for the Collaborative and Adaptive Sensing of the Atmosphere (CASA). In this CASA testbed, radars are controlled in a close-loop, collaborative and adaptive mode. The system has a fairly high temporal update rate around one minute. Case studies are presented to demonstrate the nowcasting performance for various storm types.