An examination of the logic within WFO software applications used to generate tropical cyclone impact graphics

The National Weather Service (NWS) is developing graphics to depict the potential impact of various hazards associated with tropical cyclones. These experimental graphics are currently produced by coastal Weather Forecast Offices (WFOs) upon the issuance of a Tropical Cyclone Watch or Warning within their area of responsibility. The hazards include high winds, coastal flooding (storm surge and tide), inland flooding, and tornadoes. The graphics are updated shortly after each official advisory is issued by the National Hurricane Center, and posted to NWS Web pages for use in decision-making. The graphics are color-coded to represent increasing levels of potential impact, based on automated threat assessment techniques which account for the forecast magnitude of the hazard along with the uncertainty of the forecast. The levels range from 0 to 5 (No Impact to Extreme Impact) with the definition of each level regionally determined. In order to promote consistency and limit manual editing, a set of software applications (e.g., SmartTools) is being developed and tested by NOAA's Global Systems Division and at select coastal WFOs. The goal is to provide forecasters with quality first-guess fields that can also be easily adjusted, if needed. These tools take advantage of both deterministic and probabilistic data, and are run on the common gridded forecast platform (e.g., Graphical Forecast Editor).

Given the growing importance of these graphics during tropical cyclone events, and their inherent utility to support critical decision making, the internal logic of these tools must be very sound. The information presented by the graphics must further clarify the official message, and offer realistic interpretations of threat in words that describe the potential impact. This is not a trivial matter. This paper will present the current and proposed logic for each of the hazard tools and show examples of the corresponding output. The operational availability of certain gridded input data will also be discussed. Recommendations concerning how local forecaster expertise should be integrated into the process will be made.