Environmental Analysis of GOES-R Convection Initiation Forecasting Algorithms

Convection initiation (CI) has been a complex forecasting problem, which due to several non-linear dynamic processes, is difficult to model. With the launch of GOES-R, a large increase in temporal resolution of infrared satellite data allows for the possibility of cloud top cooling (CTC) based CI forecasting algorithms to be used on incoming satellite products. Two different approaches have been applied for CTC based CI forecasting, including the University of Wisconsin's cloud top cooling algorithm (UWCTC) and the University of Alabama-Huntsville's satellite convection analysis and tracking algorithm (SATCAST). UWCTC and SATCAST use different masking and tracking procedures, but are founded on the same principles in that cooling in cloud tops can be used to forecast the onset of convection within 1 hour before the first 35 dBZ radar return. While the CI forecast principles and use have already been shown, no official study exists on environmental effects influencing CTC based algorithm performance. This presentation will outline an objective approach towards validation of CTC based algorithms and discuss the importance of the environmental parameters. Through this validation, algorithm performance is evaluated with respect to its pre-convective environmental variables, and a discriminant analysis filter is created. The discriminant analysis filter is applied to additional case studies that have been initially analyzed for performance. Filtered datasets are reexamined for performance using probability of detection (POD), false alarm ratio (FAR), critical success index (CSI) and Brier scores to determine if they have been improved. In addition, a principle components analysis is done to display where the greatest performance variance exists in convective variables. Preliminary results have shown SATCAST to over forecast several events. UWCTC is less likely over forecast due to its conservative cloud masking and typing algorithms. This presentation will show the use of several convective variables when applied to CTC based algorithms to help alleviate over forecasting. The presentation will also discuss which variables created the greatest variance in performance, which a nowcaster should be aware of when using CTC based algorithms, such as eroding CIN or upslope events. This oral presentation will display results from both the PCA and DA filtering techniques, and how well they worked on use of new datasets. It will also suggest reasoning as to certain variable performance based on GOES CTC detection. The results will be useful to forecasters using the new CTC based algorithms and researchers using fuzzy logic approaches to filter future CTC based datasets.