April 23, 2010, University of Virginia, Charlottesville, Virginia, USA

SIEDS'10 Paper Abstract


Paper FPM1Sim.4

Myers, Christopher (Univ. of Virginia)

Stochastic Models of Severe Weather Watches and Warnings

Scheduled for presentation during the Regular session "Modeling" (FPM1Sim), Friday, April 23, 2010, 14:45−15:15, Zehmer Conference Rm E

2010 IEEE Systems and Information Engineering Design Symposium, April 23, 2010, University of Virginia, Charlottesville, VA, USA

This information is tentative and subject to change. Compiled on March 5, 2015

Keywords Stochastic processes, decision analysis


To alert the public to the possibility of tornado (T), hail (H), or convective wind (C), the National Weather Service (NWS) issues watches (V) and warnings (W). The NWS Storm Prediction Center in Norman, Oklahoma, issues severe thunderstorm watches (SV), tornado watches (TV), and particularly dangerous situation watches (PV) when the likelihood of T, H, or C increases significantly. The Weather Forecast Offices issue severe thunderstorm warnings (SW) and tornado warnings (TW) when T, H, or C is imminent. The NWS characterizes the performance of warnings in terms of a probability of detection, false alarm rate, and average lead time, but these performance measures do not provide the information required for rational decision making under uncertainty. The NWS does not calculate performance measures for watches.

†††††††Two decision models are formulated that quantify uncertainty in severe weather alarms: a one-stage alarm for those who respond to warnings, and a two-stage alarm for those who respond to watches and warnings. The models identify all possible sequences of watches, warnings, and events and characterize the sequences in terms of chains of transition probabilities and distribution functions of the watch and warning durations (VD, WD), watch and warning lead times (VLT, WLT), watch and warning areas (VA, WA), and the time to warning from a watch (TTW). The models also provide a baseline for evaluating performance of a new radar technology based on dense networks of small radars developed by the Center for Collaborative Adaptive Sensing of the Atmosphere (CASA).

†††††††The one-stage alarm indicates that a TW preceded by and overlapping with a SW, defined as a reinforced warning (RW), predicts tornados better than either warning alone. The two-stage alarm indicates that watches rarely predict severe weather events, but that a SW provides longer lead times and has a lower false alarm rate when preceded by a watch. The models also indicate that RW and TW predict H and C with significant probabilities. Therefore from a decision makerís point of view, it is best to put aside the official NWS nomenclature, and treat watch and warning names as identification markers for transition probabilities and not indicators of the event to occur.



Technical Content © IEEE SIEDS.

This site is protected by copyright and trademark laws under US and International law.
All rights reserved. © 2002-2015 PaperCept, Inc.
Page generated 2015-03-05  07:58:07 PST  Terms of use