A Nowcasting System Using Full Physics Numerical Weather Prediction Initialized with CASA and NEXRAD Radar Data

As part of the annual spring experiment of the Engineering Research Center for Collaborative Adaptive Sensing of the Atmosphere (CASA), the Center for Analysis and Prediction of Storms (CAPS) ran frequently updated real-time high-resolution numerical forecasts initialized with CASA and NEXRAD radar data. In the period March 15 through June 15, 2010 whenever significant convective precipitation was expected in the CASA Oklahoma Testbed coverage area, the CAPS Advanced Regional Prediction System (ARPS) model with 1-km horizontal grid spacing was run to produce two-hour-long forecasts. These forecasts used all available surface and radar data, including reflectivity and radial velocity, through the ARPS 3DVAR and cloud analysis system, and an incremental analysis updating (IAU) procedure applied over a 5-min assimilation window, was employed to bring the analysis increments into the prediction model. Code optimization of the overall system and the use of 800 CPU cores allowed the production of each complete forecast in about ten minutes. New 2-hour forecasts were launched every ten minutes during the window of expected or observed precipitation, creating frequently updated forecasts on time schedules similar to typical nowcasting systems. When the forecasts are launched continuously, at any given time, there are 10 to 11 time-lagged forecasts available, creating a poor-man's ensemble that can be used to gauge forecast uncertainty. Forecast products created from the model output were made available in near-real-time on the Web and on a Warning Decision Support System-Integrated Information (WDSS-II) workstation in the Hazardous Weather Testbed (HWT) Experimental Warning Project (EWP). In the HWT, the forecast products, along with the raw radar data and high-resolution real-time wind analyses, were utilized and evaluated by CASA scientists, off-duty National Weather Service forecasters, and participating off-site Emergency Managers. This paper details the data assimilation design and forecast workflow for these products and highlights the April 2 and May 10 cases from the 2010 spring season that had significant severe weather. A sample 30 minute forecast of surface wind and composite reflectivity from May 20 is shown in the accompanying figure.