Using an Improved Object-Based Probabilistic Methodology to Optimize CAM Ensemble Design for Severe Weather Forecasting Applications

An object-based (OBPROB) technique is used to evaluate convective scale details of CAM ensemble forecasts, including convective initiation location, storm mode, and convective morphology. Objective evaluation using OBPROB also provides unique, quantitative information about storm mode probabilities and associated severe weather hazards that can be advantageous from an operational forecasting perspective. Such advantages have been demonstrated during recent HWT Spring Forecasting Experiments (SFEs). Forecaster feedback during the SFEs has motivated further development of the OBPROB methodology. New developments to be shown in this presentation include objective categorization of objects according to storm mode, filtering out stratiform objects, and separately displaying probabilities for discrete storms, multicell clusters and mesoscale organized convection. A case study from May 2, 2018, in which convection grew upscale into multiple bow echoes, is used to demonstrate the newly optimized OBPROB technique.

The updated OBPROB method is then used to compare the impacts of several aspects of CAM ensemble design over 10 retrospective cases from 2015 and 2016. The ensemble configurations to be evaluated include single vs. multi-model, single vs. multi-physics, fixed vs. stochastic physics, and 10-member vs. 40 member ensembles. While similar comparisons have also been previously conducted in terms of mesoscale precipitation and non-precipitation variables, this is one of the first such evaluations in terms of forecasting convective-scale details such as storm mode, morphology and related severe weather threats with objective probabilistic verification metrics.