SEMINAR: REMINDER -SEMINAR - AOML Seminar - August 21, 2012 - 10:00 a.m. - Prof. Nadia Pinardi - “Understanding ocean forecast uncertainties by ensemble,and super-ensemble methods”

[Aoml.Receptionist <aoml.receptionist@noaa.gov>]

AOML Seminar

 

Date:           Tuesday, August 21, 2012

Time:          10:00 a.m. – refreshments at 9:45 a.m.

Location:    AOML First floor Conference Room

Speaker:     Prof. Nadia Pinardi

      University of Bologna

Title:           “Understanding ocean forecast uncertainties by ensemble
and super-ensemble methods”

ABSTRACT: The Mediterranean Forecasting System (MFS, Pinardi and Coppini, 2010) has now been an operational short term numerical ocean prediction
system for more than a decade. Since its inception, MFS has developed diagnostic methods for the analysis of  initial conditions and model errors that give rise
to the growth of forecast errors  (Tonani et al, 2010).

MFS is an eddy permitting general circulation model and uncertanties peak at the mesoscales. The structure of the error is studied by several means, the first
being the study of the mesoscale horizontal and vertical field variance. The covariance matrix for temperature, salinity and sea level anomalies, following
Dobricic et al., (2007), will be described from the analysis of the variance of a 20 year long re-analysis data set (Adani et al., 2011), showing that the major
 uncertainty is the seasonal themocline depth and the mixed layer representation.

One of the major uncertainties giving rise to model errors is connected to the external atmospheric wind uncertainties and recently the MFS has developed
an innovative ensemble prediction method using a realistic distribution of winds from a Bayesian Hierarchical Model (BHM, Milliff et al., 2011). These winds
are then used to force ten days ensemble predictions to show the error growth and the uncertainty structure (Pinardi et al, 2011). In addition, a multi-model
super-ensemble method has been also developed where models with different physical parametrizations and data assimilation methods have been used to construct
a super-ensemble estimate of model SST 10 days forecast.  A synthesis of the known forecast and analyses errors for MFS will be offered from all these different methods.