Verification and Analysis of Probabilistic Hazards Information Guidance

The 2014, 2015, and 2016 Spring Forecast Experiments performed at the National Oceanic and Atmospheric Administration (NOAA) Hazardous Weather Testbed (HWT) included the introduction and testing of a proposed new method for using probabilistic hazards information (PHI) to issue severe weather notifications to emergency managers and the general public.  As part of the experiment, National Weather Service (NWS) forecasters were provided with a set of first-guess numerical weather guidance products intended to aid with and somewhat automate the process of issuing probabilistic warnings, particularly in high workload situations.  However, preliminary analysis from those experiments indicate that forecasters have difficulty understanding the basis and rationale behind the automated decision system.  In many instances, this led to frequent, and sometimes detrimental, adjustments to the automated information.  In an attempt to partially address this issue and identify areas of needed improvement, an analysis of the seasonal skill (probability of detection and lead-time) of PHI plumes auto-generated by the National Oceanic and Atmospheric Administration (NOAA) / Cooperative Institute for Meteorological Satellite Studies (CIMSS) probSevere model, will be presented and compared to NWS warning performance using local storm reports and object-based track identification. The goal of this study is to present forecasters with context on the automated system’s skill, or lack thereof, to better position the forecaster with the automated system to produce superior forecasts.