It is well-known that high-resolution mesoscale models can resolve local details of wind fields and thermal and moisture contrasts forced by underlying inhomogeneities of surface characteristics and topography. Nevertheless, the capabilities of conventional statically initialized mesoscale models are strongly limited by the lack of important mesoscale details in their initial conditions (I.C.) due to both insufficient observations and the model spin-up processes resulting from the dynamical and physical inconsistency between the model and I.C. In this paper, we demonstrate a Real-Time Four-Dimensional Data Assimilation (RTFDDA) method to alleviate this deficiency and improve the short-term forecast for a Utah-centered test region. The system timely collects all available synoptic and asynoptic observations and dynamically assimilates them into a nested-grid, 3-domain MM5 model with resolutions of 30/10/3.33km, using a continuous, 3-hourly cycling mode. The system has been operationally running since September 2000. This paper will present several case studies of locally-forced weather processes. Parallel experiments with and without FDDA were conducted for varying forecast lengths in order to quantitatively evaluate the impact of the RTFDDA system on the 0 - 12 hour forecast of local circulations. The results show an evident positive effect of the RTFDDA. Analyses of the forcing mechanisms and development processes of the meso-beta and -gamma scale circulations were also conducted. Effects of RTFDDA forecasts of different lengths are investigated. Finally, a statistical verification is performed for a late spring month to generalize the conclusion.