A Real-Time, Four-Dimensional Data Assimilation (RTFDDA) and forecast system built upon the fifth generation of the Pennsylvania State University/National Center for Atmospheric Research Mesoscale Model (PSU/NCAR MM5) has been developed and operational at NCAR. The triply nested model grids sit in the western United States over the Dugway Proving Ground in Utah, in which complex terrain and unique soil types pose challenge to model performance. This dynamically initialized system uses Newtonian Relaxation Method to bring the data and model to a mostly harmonious state before a forecast period begins. Compared to the conventional real-time forecast system that uses a one-time initialization, the RTFDDA system reduces the amount of time required for the spin-up process. In addition, the optimal use of all available data in the RTFDDA system enhances the capability to capture detailed mesoscale structures. In the study, the performance of both the RTFDDA and the conventional real-time forecast systems is evaluated statistically against a common set of observations. Both systems have demonstrated significant day-to-day fluctuations in the quantitative evaluation. The objective is to identify the strength/weakness of the RTFDDA system, relative to the conventional system (with cold-start). Analyses are carried out to study the performance of both systems under various types of synoptic scenarios, as identified and categorized during winter and spring seasons. The rationale and characteristics of each category as well as the verification statistics will be presented in this paper.