Effects of Increased Horizontal Resolution on Resolving Convection and Convective Initiation

In the summer of 2010, the Polarimetric Cloud Analysis and Seeding Test 3 (POLCAST3) field campaign evaluated the use of hygroscopic seeding flares for their possible application in the North Dakota weather modification program. During POLCAST3, local 3-km Weather Research and Forecasting (WRF) model runs were used to predict the timing, intensity, and distribution of convection and precipitation in the target region. The University of North Dakota's (UND) C-Band polarimetric Doppler weather radar was operated during the expected times of predicted convection. The WRF forecasts and radar observations are analyzed with the goal to determine forecast skill by comparing simulated and actual radar reflectivity coverage and intensity at 1-km height to spatially verify the model's predictions using new spatial verification methods. The National Center for Atmospheric Research (NCAR) Method for Object-Base Diagnostic Evaluation (MODE) Tool was used to spatially verify forecasts when precipitation was observed by the radar during a valid forecast time. While initial results suggested that the 3-km model runs performed moderately well in predicting convective coverage when both forecast and observed storms (objects) existed in the domain, with around 75% of cases falling within 10% areal coverage difference, a trend of significant under-forecasting is observed in convective complexes smaller than 45 km2. 1-km resolution runs are conducted on several case days where forecasts performed both well and poorly to determine if an increase in model resolution would help better resolve convective processes and increase forecast skill. Results from the case studies along with an overview of the usefulness of specific model output fields for predicting convective initiation are discussed.