Post flood analysis Demands of the Storm Surge Warning Service (SVSD) of the Netherlands Annette Zijderveld/ Martin Verlaan
Objective Bringing in coastal issues to the FEWS user days (second attempt) Demonstrating partnership Deltares RWS Connect networks (FEWS user community - NOOS ) Get you awake
Outline Task en process of storm surge warning NL (RWS task) Forecast quality of the SVSD Storm surge evaluations Model comparison and Forecast exchange
Process of storm surge warnings Early indication of storm surges 2 10 days before ECMWF Ensemble Prediction System 3-48 hrs before HW HIRLAM Meteo forecast 3-48 hrs before HW DCSM8 water level forecast 3-12 hrs before Data assimilation Evaluation Report Measurements Dike Watch Experience Man / Machine Mix Eastern Scheldt Barrier SVSD Maeslant barrier 3-12 hrs before SVSD warning system dike watch recommendation alert authorities alert emergency services advice closure barriers publicity internet
6 sectors for the Dutch coast Delfzijl Harlingen Den Helder Westholland Schelde Dordrecht warning level per sector local user takes action
SVSD storm surge forecast evaluation
SVSD forecast accuracy Accuracy of the SVSD storm surge forecast 50 45 Standard deviation 40 35 30 25 20 15 10 5 1954 1964 1965 1974 1975 1984 1985 1994 1995 2004 storm season Vlissingen Hoek van Holland Den Helder Harlingen Delfzijl
Extreme storm surge events Two major storms recently: 1 November 2006 9 November 2007 No flooding behind primary sea defense line, but damage in harbors and at dykes Great public interest, closing of Maeslant barrier world news Extra forecast evaluation for 1 November 2006, because of poor forecast accuracy 6 hours ahead
1 November 2006, Delfzijl
1 November 2006, Delfzijl
1 November 2006, Wadden Sea dykes
Evaluation forecast Delfzijl SVSD raised alarm with forecast of 4 m water level for Delfzijl Measured water level reached 4.83 cm + NAP
Evaluation forecast Delfzijl Investigations (1) Wind forecast incorrect Waterstand Delfzijl 30 400,00 25 380,00 20 360,00 340,00 15 320,00 10 300,00 5.0000 280,00 3.4000 5 260,00 Influence of small, local depression 3.2000 0 240,00 3.0000 220,00 200,00 Wind 4.8000 Wind speed direction Huibertgat Huibertgat Wind speed 4.6000 Nieuw Beerta 4.4000 4.2000 HIRLAM 4.0000 Gemet en 3.8000 HIRLAM 3.6000 Gemeten 2:30 2:40 2:50 3:00 3:10 3:20 3:30 3:40 3:50 4:00 4:10 4:20 4:30 4:40 4:50 5:00 5:10 5:20 5:30 5:40 5:50 6:00 6:10 6:20 6:30 6:40 6:50 7:00 7:10 7:20 7:30 A B C D E F 7:40 7:50 8:00 8:10 8:20 8:30 Sensitivity analysis for wind direction
Evaluation forecast Delfzijl 01-11-2006 Investigations (1) Hydrodynamic model(s) Resolution to low? Other models better?
Results of storm evaluation We need more (local) wind information to compare it to the model forecasts. New hydrodynamic model with higher resolution developed right now. Still local effects hard to forecast 6 or even 12 hour ahead, due to specific bathymetry issues and higher wind speed locally. Developing tools to predict forecast uncertainty.
Model comparison Need to compare operational storm surge models in order to have indication for forecasting situation, partly online validation available. Comparison operational Dutch models at operational RWS forecast center. NOOS model comparison carried out 2004/ 2005 (Noos: North West Shelf Operational Oceanographic System) http://www.noos.cc/
1 november 2006 DCSM model observed German model DCSM-UK wind Water level forecast Delfzijl 1 november 2006
MATROOS Starting to collect forecast information in 2003 within NOOS Germany UK Denmark Norway Belgium Netherlands
Operational forecast institutes and their models (2004) Country Institute Model Dimension Forecasts per day Germany BSH BSHmod 2D 2 Denmark DMI MIKE 21 2D 4 Norway MET.NO (DNMI) MI-POM 3D 2 Netherlands RIKZ/ KNMI DCSM 2D 4 Belgium MUMM OPTOS-CSM 2D 2 UK UKMO POL CS3 2D 1
Tide gauge stations 160 140 120 100 80 60 40 20 21 12 9 11 14 15 20 13 17 6 19 8 7 1 2 3 4 20 40 60 80 100 120 140 160 180 200 5 10 16 18 1 Borkum 2 Bremerhaven 3 Cuxhaven 4 Delfzijl 5 Den Helder 6 Esbjerg 7 Helgoland 8 Husum 9 IJmuiden 10 Immingham 11 Lowestoft 12 North Shields 13 Oostende 14 Roompot Buiten 15 Sheerness 16 Stavanger 17 Torsminde 18 Tregde 19 Vidaa 20 Vlissingen 21 Wick
Evaluation periods and parameters Total period: 01 October 2003-29 February 2004 Storm periods: 06-10-2003 till 10-10-2003 12-12-2003 till 17-12-2003 20-12-2003 till 23-12-2003 25-12-2003 till 29-12-2003 02-02-2004 till 07-02-2004 07-02-2004 till 10-02-2004 Evaluation parameter for surge forecasts: RMS= sqrt(average(residuals^2)) MEAN=average(residuals) RESIDUALS SURGE = model surge (t) observed surge (t) MODEL SURGE = model sea level model without wind forcing OBSERVED SURGE = measured sea level astronomical tide [from harmonic analysis or observations]
Questions (1) Are the models strongly calibrated for their local area of interest only? (2) How large are the differences between the models for one tide gauge station? (3) How large are the differences between normal weather situations and storm situations? (4) How large are the differences between short term and long term forecasts?
Results Whole evaluation period, all stations, no thresholds: rms on surge [m] 0,25 0,2 0,15 0,1 0,05 RMS on surge forecast: all stations - no threshold - 01-10-2003 : 29-02-2004 0 0-6h 6-12 h 12-24 h 24-48 h bsh_oper dmi_oper dnmi_oper knmi_noos ukmo_oper mean error [m] 0,1 0,09 0,08 0,07 0,06 0,05 0,04 0,03 0,02 0,01 0 MEAN on surge forecast: all stations - no threshold - 01-10-2003 : 29-02-2004 0-6h 6-12 h 12-24 h 24-48 h bsh_oper dmi_oper dnmi_oper knmi_noos ukmo_oper
Results Whole evaluation period, all stations, with thresholds: rms on surge [m] 0,4 0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 RMS on surge forecast: all stations - with threshold - 01-10-2003 : 29-02-2004 0-6h 6-12 h 12-24 h 24-48 h bsh_oper dmi_oper dnmi_oper knmi_noos ukmo_oper MEAN on surge forecast: all stations - with threshold - 0,05 01-10-2003 : 29-02-2004 mean on surge [m] 0-0,05-0,1-0,15 0-6h 6-12 h 12-24 h 24-48 h -0,2 bsh_oper dmi_oper dnmi_oper knmi_noos ukmo_oper
Conclusions NOOS comparison (1) Models slightly more calibrated for their own tide gauge stations, but less than expected. (2) Differences between models for one location smaller than between locations. This opens possibilities to use each others forecasts as back-up or for ensemble forecasts. Differences increase significantly in storm situations. (3) The operational Kalman filter of the DCSM model improves short term forecasts (6-12h) in storm situations. (4) The influence of weather models as main input source has not been investigated in this project, but it seems that the UK model takes advantage from good long-term wind forecasts.
Online model comparison Baysian model averaging (BMA) Den Helder, NOOS models + BMA
Future plans Now 5 years of NOOS forecast data in Matroos new evaluation Increase some statistic features in Matroos, and combine it with features of FEWS Integration of FEWS/ Matroos in new forecasting/ warning systems Example: forecasting system for IJssellake
Future plans Evaluation of new coastal models (DCSM 6, SWAN model) in preoperational mode for RWS Operational water quality forecast system (oil spill, algea bloom) Real-time verification nessecary to support forecaster Extending BMA method with other techniques
Questions?