The ESA DUE eSurge-Venice Project: Altimeter and Scatterometer Satellite Data to Improve the Storm Surge Forecasting
Zecchetto, Stefano1; De Biasio, Francesco2; Umgiesser, Georg3; Bajo, Marco3; Bellafiore, Debora3; Vignudelli, Stefano4; Papa, Alvise5; Donlon, Craig6
1Istituto Scienze Atmosfera e Clima, CNR, ITALY; 2CNR-ISAC, ITALY; 3CNR-ISMAR, ITALY; 4CNR-IBF, ITALY; 5ICPSM-Venice Municipality, ITALY; 6ESA-ESTEC, NETHERLANDS

On the framework of the Data User Element (DUE) program, the European Space Agency is funding a project to use altimeter Total Water Level Envelope (TWLE ) and scatterometer wind data to improve the storm surge forecasting in the Adriatic Sea and in the city of Venice. At present, the Storm Surge Level (SSL) is not always well forecasted: the major problems have been individuated in the inadequate wind forecast (Accadia et al., 2007, Zecchetto et al. 2005, 2007) used to force the Storm Surge Model (SSM), as well as in the poor sea level initial conditions.
The project, aimed to improve the SSL forecast by using satellite scatterometer data to calibrate the Numerical Weather Prediction (NWP) wind fields and the altimeter TWLE data to adjust the SSM initial conditions, is going to:

  • Select a number of Storm Surge Events occurred in the Venice lagoon in the period 1999-present day
  • Provide the available satellite Earth Observation (EO) data related to the Storm Surge Events (SEV), mainly satellite winds and altimeter data, as well as all the available in-situ data and model forecasts
  • Provide a demonstration Near Real Time service of EO data products and services in support of operational and experimental forecasting and warning services
  • Run a number of re-analysis cases, both for historical and contemporary storm surge events, to demonstrate the usefulness of EO data.

    The re-analysis experiments, based on hindcasts performed by the finite element 2-D oceanographic model SHYFEM (https://sites.google.com/site/shyfem/) (Bajo et al. 2097, 2010), is going to:

  • 1. use different forcing wind fields (calibrated and not calibrated with satellite wind data) (Cavaleri et al., 2004, Bertotti et al., 2009, Zampato et al., 2007);
  • 2. use Storm Surge Model initial conditions determined from altimeter TWLE data (Cipollini et al. 2008, Bouffard et al., 2008, Vignudelli et al., 2011).

    Test #1, designed to check the importance of the wind forcing on SSL evaluation, is depending on the availability of satellite wind field data.
    Test #2, designed to understand the improvements of adjusting the SSM initial conditions with TWLE altimeter data, is depending on the altimeter TWLE accuracy. The experience gained working with scatterometer and NWP winds in the Adriatic Sea (Accadia et al., 2007, Zecchetto et at., 2005, 2007) tells us that the bias NWP-Scatt wind is negative and spatially and temporally not uniform. In particular, a well established point is that the bias is higher close to coasts then offshore. Therefore, NWP wind speed calibration will be carried out on each single grid point in the Adriatic Sea domain over the period of a Storm Surge Event, taking into account of existing published methods (Cavaleri et al., 2004, Bertotti et al., 2009). The effectiveness of the wind fields calibration will be judged on the improvements in the SSL hindcast.

    The second experiment considers two different methodologies to be used in re-analysis tests. One is based on the use of the TWLE values from altimeter data in the SSM, applying data assimilation methodologies and trying to optimize the initial conditions of the simulation. In the assessment of re-analysis experiments phase, it will be checked if the uncertainties in the TWLE data from altimeter are compatible with the model sensitivity. The second possibility is an indirect exploitation of the TWLE data from altimeter in an ensemble-like framework, obtained by slight variations of the external forcing. In this case the wind data from NWP models will be weakly altered (shifted in phase), the drag coefficient will be modified, and the initial condition of the model slightly shifted in time to account for the uncertainty of these factors. Considering the ensemble of simulations, the altimeter measurements will be used to assess the RMS error of the observed and modelled data. The choice of the best configuration will be achieved with a price algorithm that explores the phase space randomly and chooses the simulation with the lowest RMS error. Starting from this simulation set up, the final forecast will be performed. The assessment will consider a possible comparison between the two types of experiments, allowing the quantification of impact of both methodologies on the final results.

    This contribution will illustrate the geophysical context of work and outline the results.