The D PHASE Operations Period (DOP)

Transcription

1 Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss The D PHASE Operations Period (DOP) Mathias W Rotach, Marco Arpagaus MeteoSwiss, Zürich, Switzerland Manfred Dorninger, University of Vienna Christoph Hegg, WSL, Zurich Switzerland Andrea Montani, ARPA SIM, Bologna, Italy Roberto Ranzi, University of Brescia, Italy Francois Bouttier, Andrea Buzzi, Urs Germann, Simon Jaun, Ken Mylne, Evelyne Richard, Andrea Rossa, Michael Staudinger, Reinhold Steinacker, Hans Volkert, Volker Wulfmeyer, Massimiliano Zappa 3 rd HEPEX Workshop, Stresa (I), June 2007

2 Outline Quick review of D PHASE The pieces that constitute D PHASE Visualisation Platform & Data Archive Conclusions DOP = D PHASE Operations Period (June 1 to November ) 2

3 MAP: Mesoscale Alpine Programme Mesoscale Alpine Programme: MAP SOP 1999 WWRP R&D Programme the first! Tons of exciting research results WWRP: output for operational forecast? > Forecast Demonstration Project (FDP) D PHASE endorsed as FDP: October

4 D PHASE Forecast Demonstration Project Theme: Heavy precipitation and flood forecast D PHASE: Demonstration of Probabilistic Hydrological and Atmospheric Simulation of flood Events in the Alpine region 4

5 Heavy precipitation & related flooding PLANAT, BWG, Luftwaffe 5

6 Distributed, Real time end to end forecasting system D Probabilistic, P high resolution, hydrological, atmospheric simulations time A H Event S E 6

7 The elements of D PHASE Atmospheric Ensemble Models High resolution deterministic models Hydrological models > including ensemble hydrological forecasts Catchments Nowcasting instruments 7

8 Probabilistic atmospheric models Model Members Resolution [km] Institution COSMO LEPS ARPA SIM MOGREPS UK MetOffice SREPS INM COSMO SREPS ARPA SIM ALADIN LAEF ZAMG PEPS variable 7 SRNWP Micro PEPS variable 2 DWD 8

9 Atmospheric models: high resolution Model COSMO COSMO COSMO COSMO MOLOCH BOLAM / MOLOCH QBOLAM ALADIN / AROME MM5 MM5 ALADIN GEM LAM Resolution 7 / 2.8 km 7 / 2.8 km 7 / 2.8 km 7 / 2.2 km 2.2 km 7 km 33 / 11 km 10 / 4 km 9 / 3 / 1 km 60 / 15 / 3.75 km 9.6 km 9 / 15 / 2.5 km Institution DWD CNMCA ARPA SIM MeteoSwiss ISAC CNR ARPA Liguria ARPAT Météo France Unv. Hohenheim FZK, IMK, IFU ZAMG Env. Canada 9

10 Atmospheric models: high resolution and others 10

11 Hydrological models > some probabilistic > e.g., based on COSMO LEPS > one driven by a radar ensemble 59 chatchments > End Users 11

12 Runoff Forecast 12

13 Ensemble hydrological forecast Various experimental systems in D PHASE Coupling local (atmospheric) ensemble system to runoff model Directly fed by radar ensemble 13

14 HEPS 1. (ECMWF EPS) 2. COSMO LEPS 3. Hydrological EPS global resolution: km ensemble members: vertical levels operational: 51 members ECMWF central and southern Europe resolution: km ensemble members: vertical levels operational: 16 members A. Montani, ARPA SIM A. Walser, MeteoSwiss PREVAH Rhine basin (Rheinfelden) resolution: km ensemble members: 51 operational: 1 member M. Verbunt, ETH Zurich S. Jaun, ETH Zurich M. Zappa, WSL 14

15 Examples probabilistic forecast Aare Hagneck (5127 km 2 ) Reuss Luzern (2251 km 2 ) q25 q75 interval q10 q90 interval min max interval observed runoff Rhine Rheinfelden (34,550 km 2 ) Rhine Domat Ems (3229 km 2 ) 15 Verbunt (2005)

16 Deterministic vs. probabilistic forecast Runoff at Hagneck (Aare, 5170 km 2 ), IT: , 00 UTC Different determinstic runs > dependent on initial time S. Jaun, IACETH Observed runoff Inter quartile range of the ensemble 16

17 Radar Ensemble for hydrological modeling Use estimation of radar uncertainty > shading by topography > clutter error covariance matrix construct an ensemble of radar precipitation estimates Germann et al, ERAD, 2006 (keynote) Germann et al, EGU, 2007 Germann et al, AMS Radar Conf,

18 Radar Ensemble for hydrological modeling 2864m FOEN river gauges Pincascia Lavertezzo 44km 2 193m Verzasca Lavertezzo 186km 2 10 km FOEN = Swiss Federal Office for the Environment automatic raingauges In collaboration with M Zappa (WSL), M Berenguer (Montreal), I Zawadzki (Montreal), 18 D Sempere Torres (Barcelona).

19 Nowcast run 14 May 2007 (without Frasco) Radar Ensemble River Radar ensemble Germann et al, ERAD, 2006 (keynote) Germann et al, EGU, 2007 Germann et al, AMS Radar Conf, m 3 /s 25 m 3 /s Runoff observed runoff Hydrological model based on rain gauge Time hrs 19

20 Nowcast run 15 May 2007 (with Frasco) Radar Ensemble River Radar ensemble Germann et al, ERAD, 2006 (keynote) Germann et al, EGU, 2007 Germann et al, AMS Radar Conf, m 3 /s 50 m 3 /s observed runoff Runoff Hydrological model based on rain gauge Time hrs 20

21 Nowcasting Instruments Tool Domain Institution VERA D PHASE / COPS Domain University of Vienna NWP minus VERA D PHASE / COPS Domain University of Vienna CLEPS minus Satellite D PHASE Domain DLR NASS (quantitative precipitation estimate based on radar) Piemonte MeteoSwiss radar Composite Switzerland Switzerland / Piedmont MeteoSwiss ARPA SIM / MeteoSwiss ENASS Switzerland MeteoSwiss TRT (Thunderstorm tracking tool) Switzerland MeteoSwiss 21

22 Common (centralised) Platforms Visualisation Platform (VP): to visualise (display) all products and all alerts; main source of information for forecasters and end users interface for (some of) the users to submit their feedback on the available products and alerts, respectively realised by Next Generation Software (Salzburg) phase.info > pw required 22

23 VP, Level 1: Alpine region model products Nowcasting products feedback select forecast range select accumulation time go to: monitoring products feedback

24 VP, Level 2: Alpine sub region, country model products nowcasting products monitoring products feedback

25 VP, Level 3 a: target area Same alerts for all models > same thresholds > same software

26 VP, Level 2: Alpine sub region, country today 44% 57% model products nowcasting products monitoring products feedback

27 VP, Level VP, Level 3 h: 3 a: impact target area area (catchment)

28 VP, Level 1: Alpine region model products Nowcasting products feedback select forecast range select accumulation time Same plots for all models > same variables > same color coding go to: monitoring products feedback

29 Joint visualisation thanks to COPS 15 min accumulated precipitation

30 Joint visualisation thanks to COPS Relative humidity, 850 hpa wind

31 Common (centralised) platforms Data Archive (DA) to archive all data, alerts, and feedbacks; mainly for verification purposes also stores all COPS and GOP data physically based at the MPI in Hamburg 31

32 Summary End to end modelling system who s who in the high resolution / probabilistic modelling world Probabilistic modelling of hydrological processes Large number of end users Data set for model verification / feedback evaluation 32

33 MAP History Thank you for your attention! 33

34 Feedback by Forecasters Helped to design Visualisation Platform Evaluation by forecasters > daily > weather type, type of event,. > what did I use? > what did support me? Analogue feedback by end users 34

35 Feedback by Forecasters today 44% 57% model products nowcasting products monitoring products feedback 35

36 Feedback by Forecasters Helped to design Visualisation Platform Evaluation by forecasters > daily > weather type, type of event,. > what did I use? > what did support me? 36

37 Additional guidance to Forecasters Alerts from COSMO LEPS: probabilities Additionally: return periods If the model has extreme precipitation > 100mm/24h, say: > is it an extreme model event? > concept based on Extreme Forecast Index (EFI) Model climatology > for each day > seasonality and location > daily updated > on the VP 37

38 Return periods COSMO LEPS 15 days 24h Niederschlag 2m Temperatur 10m Wind 30 days Wiederkehrperiode statt EFI Probability for 24h precipitation sum to be larger than the n day maximum in the model climatology 60 days 180 days 38

39 Analysis and verification Prior to D PHASE > model testing, sensitivity, etc. > MAP cases (and others) Real time evaluation > VERA (University of Vienna) Analysis of interesting cases > all participants > post festum Objective vs. subjective verification > post festum 39

40 More details: Analysis for each model Example: ALADIN model by ZAMG, Austria Precipitation sum averaged over all impact regions Fraction of impact regions covered by the model Temporal coverage (RR mod RR obs ) / RR obs Mean over all impact regions Mean absolute value in one impact region Equitable threat score, threshold 1mm/h Correlation of hourly time series for each impact region (Results of other models: see Appendix) 40

41 QPF Verification Summary JJA + + Overall over / underestimation? Ability to resolve mean spatial patterns. Timing 41

42 QPF Verification Summary June + + Overall over / underestimation? Ability to resolve mean spatial patterns. Timing 42

43 QPF Verification Summary July + + Overall over / underestimation? Ability to resolve mean spatial patterns. Timing 43

44 QPF Verification Summary August + + Overall over / underestimation? Ability to resolve mean spatial patterns. Timing 44

45 Alert Verification: Level yellow Alert level yellow (10 events per year), aggregated on 6 hour intervals. probability to detect a real alert event frequency model alert / frequency of real alert Dependency on accumulation period: 03h 06h 12h 24h 48h 45

46 Alert Verification: Level orange Alert level orange (2 events per year), aggregated on 6 hour intervals. probability to detect a real alert event frequency model alert / frequency of real alert Dependency on accumulation period: 03h 06h 12h 24h 48h 46

47 Summary End to end modelling system who s who in the high resolution / probabilistic modelling world Probabilistic modelling of hydrological processes Large number of end users Data set for model verification / feedback evaluation 47

48 MAP History Thank you for your attention! 48