Modelled and observed winds in the Adriatic sea F. De Biasio, S.Zecchetto, A. della Valle ISAC-CNR Institute of Atmospheric Sciences and Climate Italian National Research Council http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 1
Presentation outline Storm Surge Networking Forum 2013 Motivations Operational model winds in the Adriatic Sea Operational satellite winds in the Adriatic Sea Verify model winds against satellite winds Modify model winds according with satellite data Conclusions http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 2
Motivations The surge in the Adriatic Sea is driven by several phenomena, but the most important are wind and pressure*. In order to forecast the surge, we need atmospheric forcing forecast. The wind is routinely observed by scatterometers: satellite winds can be used to verify/ect model winds *Lionello et al., 2006, Roland et al., 2009 GFS model - Mittel-Europa Forecast downloaded from Wetterzentrale.de http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 3
Model winds in the Adriatic Sea at ICPSM GLOBAL MODELS: IFS operational at ECMWF From 21/11/2000 to 01/02/2006: From 01/02/2006 to 26/01/2010: From 26/01/2010: Grid spacing used (after interpolation): TL511 (40 km) TL799 (25 km) TL1279 (16 km) 12,5 km LOCAL AREA MODELS: ALADIN operational at DHZ (Croatian Met. and Hydrol. Service) initialized with ARPEGE (Meteo-France) Grid spacing: 8 km COSMO-LAMI operational at ARPA Emilia-Romagna initialized with ECMWF, one grid Grid spacing: 7 km http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 4
Satellite winds in the Adriatic Sea SENSOR: SeaWinds on QuikSCAT ASCAT on MetOp-[A B] OSCAT on OCEANSAT-2 TYPE: Ku-band pencil beam scatt. C-band scatterometer Ku-band pencil beam scatt. AVAILABILITY: 1999-2009 2009-present 2010-present GRID SPACING: 12.5 km 12.5 km 12.5 km SWATH: One swath 1800 km wide Two swaths 550 km wide One swoth 1800 km wide AVERAGE HITS: ~30/mo in the Adriatic Sea ~20/mo in the Adriatic Sea ~30/mo in the Adriatic Sea DIST. TO COAST: 20 km 35 km std, 15 km coasta 20 km Coverage: Data downloaded from PO.DAAC http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 5
Verify NWP model winds against satellite winds We need to know how much the forecast differ from observations... A Bora event (2008/09/21) http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 6
Verify NWP model winds against satellite winds We need to know how much the forecast differ from observations... A Sirocco event (2008/11/04) http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 7
Verify NWP model winds against satellite winds We need to know how much the forecast differ from observations... Results: ECMWF has better overall performances http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 8
Verify ECMWF model winds against satellite winds We need to know how much the forecast differ from observations, and how: systematic discrepancies NWP models underestimate the wind http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 9
Verify NWP model winds against EO data We need to know how much the forecast differ from observations, and how: variability in space Δws = ws scatt ws ws scatt Δθ = θ scatt θ http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 10
Verify ECMWF model winds against satellite winds We need to know how much the forecast differ from observations, and how: variability in time Δws = ws scatt ws ws scatt Δθ = θ scatt θ http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 11
Modify model winds according with satellite data Three prescriptions: The model forecast wind fields could be adjusted using the found biases The ection should reflect the space variability The ection should reflect the temporal variability computation of the mean of SCATT-ECMWF normalised wind speed difference ΔWS / WS scat and wind direction difference Δθ relative to scatterometer direction over a 7 days period around the SEV occurrence ws = ws (1 + Δws) θ = θ + Δθ http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 12
Modify model winds according with satellite data Three prescriptions: The model forecast wind fields could be adjusted using the found biases The ection should reflect the space variability The ection should reflect the temporal variability ws = θ = θ ws ( x, ( x, y) y) http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 13
Modify model winds according with satellite data Three prescriptions: The model forecast wind fields could be adjusted using the found biases The ection should reflect the space variability The ection should reflect the temporal variability ws = θ = θ ws ( x, y, t) ( x, y, t) http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 14
Conclusions no SSM dependence on the model wind grid spacing (50 km 16 km); with tuned model winds the SSM peak error drops to 6 cm from 10 cm; RMSE decreses by 8%, Correlation rises by 1%; Yet... the representativeness of the relative bias (few scatt data in the SEV window) has to be assessed: longer windows allow more stable statistics but make the dynamic tuning less effective; the dependence of the relative bias on meteorological conditions is not totally clarified; need of a longer dataset to verify the tuning algorithm. http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 15
Thank you for your attention http://www.esurge-venice.eu CNR-ISMAR, 18-20 November 2013 16