Open session on Operational Oceanography -OS4.1/CL4.11 EGU Use of ensemble forecast meteorological fields to force a storm surge model Marco Bajo, Luciana Bertotti and Luigi Cavaleri ISMAR-CNR, Venice, Italy
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Summary The problem: Venice flooding; The method: Ensemble forecast forcing; Case I: 4 November 1966 (the highest flood); Case II: 1 December 2008 (operational forcing); Conclusions.
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Venice location Venice lagoon City of Venice
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Venice flooding Even a small storm surge (e.g. 30cm) with the tidal contribution can cause the flooding of the City. Sea levelFlooded surface 190 cm100% 140 cm90% 130 cm70% 120 cm35% 110 cm12% 100 cm4% Pavement lower than 90 cm
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Storm surge forecast in Venice: ICPSM Both statistical and dynamical models are used for the storm surge forecast Operational: 24/24 hours - 7/7 days After the 1966 event an Office for the sea level forecasting and warning has been created (ICPSM)
Open session on Operational Oceanography -OS4.1/CL4.11 EGU ECMWF wind and pressure forecast Tide gauges Models Subjective forecast Alerting population Other observations Operational storm surge forecast in Venice ICPSM operational chain
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Hydrodynamic model SHYFEM (Shallow water HYdrodynamic Finite Elements Model) Model main features: - 2D shallow water equations; - Staggered finite elements scheme; - Semi-implicit time discretisation scheme; - Computational domain: Mediterranean Sea; - Open boundary in Gibraltar. Level set to 0, free normal fluxes. Forcing: - ECMWF wind and pressure forecasts. Open source code. Visit: And
Open session on Operational Oceanography -OS4.1/CL4.11 EGU The ECMWF ensemble meteorological forecast Uncertaintes in the initial state Phase space DET ENS Time ensemble members; - Perturbations of the initial conditions (singular vectors); - Model perturbations.
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Method Ensemble low res. forecasts 1 High res. deterministic forecast Simulation of 2 storm surge events: 4 November December 2008 Modeled surge is extracted near Venice and compared with observations (without the tidal signal) FORCING Ensemble runs 1 Deterministic run SEA LEVEL FORC.
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Case I: 4 November November 1966, Venice (from The highest event ever measured (1.94 m, with a storm surge of 1.8 m); City flooded for a long time, more than 30 hours; No operational forecast available.
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Case I: 4 November 1966 run: 2 November 1966, 00UTC Time shift due to different locations of the data Time variation of the ens. peaks 22 hours seiches Ens. divergence
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Case I: 4 November 1966 Errors on the estimation of the maximum IQR =0.11 m ENS =0.51 m C =0.58 m DET = 0.49 m 5days 4days 3days 2days 1day Error [m] No relation A posteriori A priori
Open session on Operational Oceanography -OS4.1/CL4.11 EGU December 2008, Lido beach, Venice Case II: 1 December 2008 Highest event of the last 20 years; Storm surge peak had the same time of the tidal peak; The storm surge event was not isolated.
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Case II: 1 December 2008 run: 29 November 2008, 00UTC Good estimation Low time or spatial resolution? 22 hours seiches Seiches + forcing
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Case II: 1 December 2008 Errors on the estimation of the maximum IQR =0.15 m ENS =0.12 m C =0.17 m DET = 0.24 m 5days 4days 3days 2days 1day Error [m]
Open session on Operational Oceanography -OS4.1/CL4.11 EGU Conclusions - The ensemble median can give an accurate estimation of the maximum storm surge peak, more coherent between different runs; - Possible estimate of the forecast error considering the ensemble width (e.g., IQR, standard deviation); Possible improvements: -Increase the temporal and spatial resolution of the forcing;