P2.7 A QUANTITATIVE EVALUATION ON THE PERFORMANCE OF A REAL-TIME MESOSCALE FDDA AND FORECASTING SYSTEM UNDER DIFFERENT SYNOPTIC SITUATIONS RONG-SHYANG SHEU*, JENNIFER CRAM, YUBAO LIU, AND SIMON LOW-NAM National Center for Atmospheric Research Boulder, Colorado 1. INTRODUCTION A mesoscale weather analysis and forecasting system that employs Real-Time Four-Dimensional Data Assimilation (RT-FDDA) has been developed and operational since summer of 2000. The forecast system is built upon the Fifth Generation of the Pennsylvania State University/National Center for Atmospheric Research (PSU/NCAR) Mesoscale Model (MM5). The triply nested model domain sits on the western United States and centers on the Dugway Proving Ground in Utah, where complex terrain and surface characteristics account for a variety of local forcing. The system utilizes a continuous data ingest mechanism to produce an analysis period using Newtonian relaxation method, before a forecast period begins. The goal of the pre- forecast analysis period is to bring the model to a mostly dynamically balanced state so as to reduce the time required for the spin-up process during the forecast period. A detailed description of the system can be found in this volume of preprints (Cram et al., 2001). The RT-FDDA system is run in parallel to an operational real-time forecast system (referred to as OPN hereafter) that uses a one-time, static initialization at the beginning of the forecast [Davis et al., 1999]. Both systems are validated routinely against observations. Both sets (OPN and RT-FDDA) of verification statistics display significant fluctuations on the time scale of 3 to 5 days. Examples of this feature are shown in Figure 1, in which temperature bias at 1800 UTC (panel a) and 0000UTC (panel b) are plotted as a function of day in January 2001. The statistics are generated based on grid 3 solutions. It is also noted that the two sets of validation statistics are more diverged on some days than others. Assuming the quality of observational data holds steady, the synoptic situations could be reflected in the day-to- day fluctuations of statistics. Hence, the objectives of this study include (a) to identify how the performance of the forecast systems, in terms of verification statistics, are modulated by different synoptic scenarios, and (b) to identify the relative strength of the RT-FDDA forecast system to the regular cold-start operational system. 2. MODEL CONFIGURATION/PHYSICAL OPTIONS * Corresponding author address: Rong-Shyang Sheu, NCAR/RAP, P.O. Box 3000, Boulder, CO 80307; e-mail: sheu@ucar.edu. The OPN system has undergone a series of changes and upgrade over the years since it became operational. See Davis et al. (1999) for detailed description of the original configuration. The current system preserves most of the configuration and physical options, with the most significant upgrade/changes being 1) first-guess background fields for the initial conditions coming from new NCEP Eta model forecasts of finer resolution; 2) using the Oregon State University Land Surface Model (OSU LSM) for the calculation of substrate temperature and moisture (Chen and Dudhia, 2001). The coarse grid resolution is 30 km, with three finer grids at 10 km, 3.3 km, and 1.1 km resolution. The system produces two 36-hour forecasts that begin at 0600Z and 1200Z daily. -5 0 5 T Bias (˚C) (a) RTFDDA OPN -5 0 5 T Bias (˚C) 5 10 15 20 25 30 Day of Month (b) Fig. 1, Examples of temperature forecast bias fluctuating as a function of day in January 2001. Panel (a) shows verification results valid at 1800 UTC; whereas Panel (b) shows verification results valid at 0000 UTC. The solid lines represent the results from the RT-FDDA forecasts with beginning time valid at 1300 UTC; the dashed lines represent results from OPN forecasts. The RT-FDDA system is configured to use mostly the same physical options as those adopted by the OPN system. The major difference, other than the dynamic