1 / 15

The challenge to verify operational weather warnings

The challenge to verify operational weather warnings. Tanja Weusthoff and Marco Arpagaus EMS, 14.09.2011. MeteoSwiss official warnings (24h acc. precip.) for 4th September 2011. 152 warning regions. Introduction. 5 level. 8 different hazards.

donny
Télécharger la présentation

The challenge to verify operational weather warnings

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. The challenge to verify operational weather warnings Tanja Weusthoff and Marco Arpagaus EMS, 14.09.2011 MeteoSwiss official warnings (24h acc. precip.) for 4th September 2011

  2. 152 warning regions Introduction 5 level 8 different hazards  currentproject: developmentof an applicationfor an automaticverificationofweatherwarnings

  3. Introduction specifications: • develop an automaticverificationforoperational warnings, whichallows a certaintolerance (spaceand time) • usesynergies(e.g. with GIN, a commonplatformfor all naturalhazards in Switzerland: http://www.gin-info.ch/index.html) motivation: • replacecurrentmanual(subjective) verification in ordertoreleaseresources • provethequalityofofficialwarnings • singleofficialvoice (SOV, sinceJanuary 2011)  distribution via mediaforlevel 4 and 5

  4. Challenges • how to evaluate the usefulness of warnings without knowing the needs of individual users (and their cost/loss) • how to interprete „tolerant results“ and what should be communicated to the users Precautions causes Costs Having no protection results in Losses

  5. Challenges • representativityofobservations • accountforfeedbackofauthorities • smallstatistics (rare events) e.g. station Magadino / Cadenazzo (203 m asl) is only wind station for two warn regions with complex terrain (308 and 309) > 2000 m < 800 m

  6. Basic Concept event-oriented verification, warnings as binary events tolerant in space, time and threshold distinguish two types of warnings SHORT: short-term events (e.g. thunderstorms) AKKU: accumulated events (e.g. 24h precipitation accumulation) event definition and evaluation is (in principle) the same within each group; differentiate between basic verification (strict) and detailed verification verification per warning region, summary for specific regions or whole Switzerland

  7. Example: AKKU Rain  level 3-5 Snowfall (lowlands and mountains)  level 3-5 Snowmelt  level 3-5 Heat wave level 3 • snowmelt cannot be verified due to a lack of observations; • for snow and rain use of radar data (spatial information) and psychrometer temperature

  8. Event-Definition WarnEvent (ts to te) acc/3 ti issue time t2 te ts t1 ObsEvent (t1 to t2) t2 - ti > tv? AKKU • each warning is an event; duration of warning at least as long as accumulation period duration • observation: consider hourly 24h, 48h and 72h sum; important is first threshold exceedance (t2) Hit  An eventisobservedduring a valid warningandthefirstthresholdexceedance (t2) occursat least accumulationPeriod/3 hours (i.e. 8h,16h,24h) after thebeginningofthe warn event (ts) and not laterthenthe end ofthe warn event (te). Miss  A thresholdexceedanceisobservedwithout an activewarningorthefirstthresholdexceedanceoccurs bevor accumulationPeriod/3 hoursafter thebeginningofthewarning (ts). FalseAlarm  A warninghasbeenissued, but nothresholdexceedancehasbeenobserved.

  9. Evaluation A: Basic-verification Present results as: POD FAR FBI (TS)  derived from contingency table AKKU Missing-D-problem: what is a „non-event“? Evaluation B: detailed verification • in principle like Evaluation A (Hit, Miss, False Alarm) • introduce additional category • combination of Hit, Miss und False Alarm  see definitions on next slide

  10. Evaluation B: detailed verification WarnEvent (ts to te) ti issue time te t2 ts t1 ObsEvent (t1 to t2) t2 - te < tshift AKKU (2.) (1.) WarnEvent (ts to te) acc/4 acc/3 ti issue time t2 te ts ObsEvent (t1 to t2) Specifications Miss + Hit + False Alarm  the first threshold exceedance (t2) occurs maximal tshift hours after the end of the warn event (te) or less than accumulatioPeriod/3 (8,16,24h) but more than accumulationPeriod/4 (6,12,18h) after the start of the warning (ts).

  11. Evaluation B: detailed verification Presentation of results: „perfect“ hit „useful“ combined categories including a hit „bad“ false alarm, miss AKKU „good“ ... adapted from DWD

  12. Tolerance Allowtolerances in: threshold:LowHit = 90% ofthreshold time:variationoftime components (e.g. tshift) space:ifpossible, takeintoaccountneighbouring warn regions applytolerancestobasicanddetailedverification SHORT: possibilitytorequest a minimumleadtimetvforthewarning (i.e. a hitisonlypossibleifthewarninghasbeenissuedat least tv (= t1 - ti) hoursbeforetheobservedevent, otherwiseitisclassifiedas a miss) SHORT + AKKU

  13. Application  Flow Diagram warn DataBase  warn events • Java Application • AWV • deriveobsevents • perform event-basedverificationforeachsettingandeachregion store warn and obs events on a monthly basis obs DataBase  hourly data per warn region store evaluation results for each event on a monthly basis • aggregateresultsandcalculatescoresforspecificperiod • present results • per region • single event • …

  14. Luzern MeteoSwiss official warnings (24h acc. precip.) for 4th September 2011 Observations (surface stations): 24 h acc. precipitation analysis for 4th September 2011 (prel.) Bern A qualitative example… Level 3 warning: WarnIssue 04.09.2011 08:56 WarnStart 04.09.2011 12:00 WarnEnd 05.09.2011 12:00 24h acc. precip. 35 mm/24h 50 mm/24h

  15. Thanks for listening …

More Related