NWC REU 2021
May 24 - July 30



Photo of author

Tornado Warning Decision Making with the Probabilistic Hazard Information Tool

Kerstin L. Gillespie, Daphne S. LaDue, and Christopher D. Karstens


What is already known:

  • NWS forecasters currently issue storm-based warning polygons when they determine that a tornado is imminent or present.
  • These warnings are significantly larger than the threat, generally encompassing the entire storm.
  • Much of the area in the warning does not experience severe weather conditions, causing people to lose trust in the system.
  • Forecasters are limited in how they can update these warning polygons before they must issue a new one.
  • The prototype Probabilistic Hazard Information tool, tested from 2014–2017, seeks to reduce these problems by allowing forecasters to narrow down the threat area and express their confidence on where and when a tornado might occur.

What this study adds:

  • Three different forecasters each issued experimental PHI “warnings” on the same two storm cases, talking aloud as they worked.
  • This talk-aloud data reveal that forecasters were able to anticipate tornado formation for these cases; they did not have confidence to project maintenance beyond about 20 min.
  • The PHI tool helped forecasters express more of their thinking, including doubt that a tornado threat existed during times a storm was cycling or reorganizing.


The National Weather Service’s (NWS) current warning system has been in place for several decades. Research has shown it has limitations with more precisely depicting the area of tornado threats, and the inability to update the warnings without reissuance, among others. This has motivated scientists to develop new ideas for warning systems, one of which is called the prototype Probabilistic Hazard Information project. Research done on this tool shows that it has limitations, but also the potential for adding to or replacing the current warning paradigm. Nine NWS forecasters were brought into the NOAA Hazardous Weather Testbed in 2017 to test the tool and warning strategy by working through several weather cases. For this study, three cases of talk-aloud data from the forecaster issuing tornado PHI objects were analyzed to determine how they made tornado warning and advisory decisions using the new tool. Findings show that forecasters were able to use the tool to accurately predict tornado formation. In contrast, tornado maintenance was not anticipated as accurately, with the tool showing their confidence decreasing over the duration of time that each PHI object would be in effect (usually 60 min). These findings indicate that PHI is useful for conveying the prediction of tornado formation and shows promise for future use in the NWS warning paradigm.

Full Paper [PDF]