Evaluation of a rapidly relocatable high-resolution numerical model for meteorological nowcasting based on MM5
Nelson L. Seaman, Penn State Univ, University Park, PA; and D. R. Stauffer, A. Deng, A. M. Gibbs, A. J. Schroeder, and G. K. Hunter
Short-term weather forecasts and nowcasts are increasingly important to the economy (e.g., commodity trading, retain sales, travel) and personal safety (accidental or intentional release of chemical, biological or nuclear materials). Although it is sometimes important to have meteorological information over very large regions, it is now common to need very fine-scale nowcast guidance over specific areas of 1-5 km or even smaller. How to best provide accurate short-term highly localized predictions is a vital area of research because it is generally unacceptable to merely use existing models designed for mesoalpha-scale applications at much finer scales. The internal physical parameterizations may not be readily adaptable to finer grids. Moreover, traditional static initializations of synoptic and mesoscale forecasts can require several hours to approach equilibrium states. If the time horizon of interest is 1-3 h, then the dynamic adjustments during this "spin-up" period can severely damage the accuracy of solutions. Clearly, considerable work must be done to provide more accurate fine-scale short-term forecasts.
An optimized full-physics version of the PSU/NCAR MM5 has been designed and built that is suitable for a variety of fast-response nowcast applications. The model is configured with triply nested grids of 36-km, 12-km and 4-km, each with 30 layers in the vertical. The inner 4-km domain covers an area of 500 X 500 km. The model is globally relocatable in about 5 minutes and runs in real time on a dual processor 933-MHz PC, producing updated meteorological fields every 30 minutes. Data ingest is currently from Unidata and a global model is used for initial and boundary conditions. Sophisticated scripts allows the system to be operated by a user with no meteorological or numerical training. Simple adaptations of this numerical nowcasting system are being developed for the U.S. Army, the MA Dept. of Public Health and other users.
The MM5 nowcaster model has been applied in two special studies of about three weeks duration each. The first field test was conducted over OK in August 2001 and the second was held in the Middle Atlantic states in March 2002. Statistical analysis of these field tests provides insights into model performance from the surface layer to the stratosphere. Case study analyses highlight model performance in situations having special circumstances, such as deep convection or frontal boundaries. Results from these early evaluations supply information for future model improvements in terms of improved physics and dynamic initializations.
Extended Abstract (272K)
Joint Session 6, The Use of Mesoscale Models in Severe Local Storms Forecasting (Joint with 21SLS and 19WAF/15NWP)
Tuesday, 13 August 2002, 11:15 AM-12:00 PM
Previous paper Next paper
Browse or search entire meeting
AMS Home Page