The NOAA/CIMSS ProbSevere Model: Incorporation of Total Lightning and Validation

Abstract

The empirical Probability of Severe (ProbSevere) model, developed by the National Oceanic and Atmospheric Administration (NOAA) and the Cooperative Institute for Meteorological Satellite Studies (CIMSS), automatically extracts information related to thunderstorm development from several data sources to produce timely, short-term, statistical forecasts of thunderstorm intensity. More specifically, ProbSevere utilizes short-term numerical weather prediction guidance (NWP), geostationary satellite, ground-based radar, and ground-based lightning data to determine the probability that convective storm cells will produce severe weather up to 90 min in the future. ProbSevere guidance, which updates approximately every 2 min, is available to National Weather Service (NWS) Weather Forecast Offices with very short latency. This paper focuses on the integration of ground-based lightning detection data into ProbSevere. In addition, a thorough validation analysis is presented. The validation analysis demonstrates that ProbSevere has slightly less skill compared to NWS severe weather warnings, but can offer greater lead time to initial hazards. Feedback from NWS users has been highly favorable, with most forecasters responding that ProbSevere increases confidence and lead time in numerous warning situations.

Adapting Satellite Soundings for Operational Forecasting within the Hazardous Weather Testbed

“…One forecaster suggested tools to provide an alert when an overpass is available in their forecast area of responsibility. Products within AWIPS that bring attention to potential hazards, such as ProbSevere [37,38] (which was also tested in the HWT), were very useful to forecasters who must analyze significant amounts of data [11]. Thus, NUCAPS could also benefit from alerting forecasters to regions of high CAPE or large gradients in the lapse rate.…”

Development of a Human–Machine Mix for Forecasting Severe Convective Events

“…To facilitate a human-machine mix for warning applications within the prototype system developed by Karstens et al (2015), a consolidation of automated guidance into an object-based framework must occur. Key components of this process include object identification (i.e., vectorizing) of atmospheric processes (i.e., severe convective storms) from continuous fields with self-describing attributes (e.g., Lakshmanan et al 2009), object tracking through time (e.g., Lakshmanan and Smith 2010;Lakshmanan et al 2015), update frequency sufficient to resolve hazard evolution (e.g., LaDue et al 2010;Heinselman et al 2012;Wilson et al 2017), and hazard prediction (e.g., Cintineo et al 2014). Three forms of automated object-based guidance were incorporated and tested with forecasters, with efforts beginning in 2015 (Table 1).…”

“…Three forms of automated object-based guidance were incorporated and tested with forecasters, with efforts beginning in 2015 (Table 1). Guidance for the creation of combined severe thunderstorm forecasts was based on the NOAA ProbSevere model (Cintineo et al 2014) via objects identified from MRMS composite reflectivity with an update frequency of approximately 2 min (Fig. 2).…”

“…The ProbSevere model combines data from Geostationary Operational Environmental Satellites (GOES; Menzel and Purdom 1994), the WSR-88D network, and the Rapid Refresh model (RAP; Benjamin et al 2006) to probabilistically forecast the severe weather likelihood from a developing storm. The ProbSevere model evolved to incorporate lightning predictors and offers probabilistic predictions for individual severe thunderstorm hazards (hail, wind, and tornado; Table 1; Cintineo et al 2018). ProbSevere guidance was also augmented with predictions of duration (McGovern et al 2017), storm classification (McGovern et al 2018), damaging straight-line winds (Lagerquist et al 2017), and probability trend prediction (Harrison et al 2018) using machine learning techniques.…”

“…These algorithm development efforts quickly evolved to include multisensorenabled algorithms using four-dimensional radar data (Lakshmanan et al 2007) with the focus shifting to automated detection of severe weather phenomena. This Multi-Radar Multi-Sensor (MRMS) system (Smith et al 2016) has been combined with satellite and environmental information to drive artificial intelligence applications (e.g., Cintineo et al 2014;McGovern et al 2017;Gagne et al 2017) and emerging convection-allowing model ensemble systems (e.g., Stensrud et al 2009;Wheatley et al 2015) that aim to predict the likelihood of severe convective weather occurrence, all with the goal of enhancing forecaster situational awareness and extending warning lead time.…”