Extended Abstract (112K)P1.5
The Determination of Optimal Thresholds of Tropical Cyclone Incremental Wind Speed Probabilities to Support Expressions of Uncertainty in Text Forecasts
Pablo Santos, NOAA/NWS, Miami, FL; and D. Sharp, M. DeMaria, and S. Kiser
As tropical cyclone events unfold, decision-makers require a meteorologist's most likely wind speed forecast, along with an accompanying expression of uncertainty. Both are necessary to effectively manage preparations for life-threatening weather events. The inherent uncertainty in tropical cyclone forecasts results in the shortcoming of deterministic-only wind speed forecasts such as those found within the current Zone Forecast Product and Coastal Waters Forecast issued by National Weather Service (NWS) Weather Forecast Offices. To address this shortcoming, the NWS offices in Miami and Melbourne have developed an experimental means to consistently and coherently incorporate uncertainty information in these products through the creative use of the National Hurricane Center's incremental wind speed probabilities. These probabilities are created upon the issuance of each official advisory for 34-, 50-, and 64-knot winds through 120 hours. The probabilities are derived using uncertainties in the official track and intensity forecast, and also from a climatology and persistence wind radii forecast model (with error statistics determined from the previous five years of forecasts). Then, the probabilities are configured locally, in gridded form, to match the traditional time increments of the textual Zone Forecast Product and Coastal Waters Forecast. Together with gridded hazard information (e.g., tropical storm/hurricane watches/warnings) and gridded deterministic wind speed information, the tropical cyclone incremental wind speed probabilities trigger enhanced wording which conveys the situational uncertainty for successive forecast periods. The logic has been encoded within tropical cyclone versions of the respective text formatters which invoke the prescribed expressions. This includes the zone-based versions which are used to generate the legacy products, and the dynamic point-n-click (point-based) versions found on Weather Forecast Office web sites. The logic depends, in part, on the exceedance of preliminary incremental wind speed probability thresholds as a function of time (e.g., forecast period) for uncertainty involving tropical storm force winds and hurricane force winds. Experimental and operational testing has yielded encouraging results. Yet, since the incremental wind speed probability thresholds are a critical component to the formatter logic, it is prudent to formally identify those values which tend to maximize the responsible detection of a potential event while minimizing false alarms which can lead to the overuse of enhanced expressions. Optimal thresholds were determined using Relative Operating Characteristics Diagrams and Peirce Skill Scores. This paper will present the validation results of the stated incremental wind speed probability thresholds study, identifying those optimal values as a function of time for coastal locations stretching across the U.S. Gulf of Mexico and Atlantic coasts.
Poster Session 1, Modeling, Forecasting, and Impacts of Urban High Impact Weather
Wednesday, 14 January 2009, 2:30 PM-4:00 PM, Room 126B
Previous paper Next paper